1 acpi= [HW,ACPI,X86,ARM64]
2 Advanced Configuration and Power Interface
3 Format: { force | on | off | strict | noirq | rsdt |
5 force -- enable ACPI if default was off
6 on -- enable ACPI but allow fallback to DT [arm64]
7 off -- disable ACPI if default was on
8 noirq -- do not use ACPI for IRQ routing
9 strict -- Be less tolerant of platforms that are not
10 strictly ACPI specification compliant.
11 rsdt -- prefer RSDT over (default) XSDT
12 copy_dsdt -- copy DSDT to memory
13 For ARM64, ONLY "acpi=off", "acpi=on" or "acpi=force"
16 See also Documentation/power/runtime_pm.rst, pci=noacpi
18 acpi_apic_instance= [ACPI, IOAPIC]
20 2: use 2nd APIC table, if available
21 1,0: use 1st APIC table
24 acpi_backlight= [HW,ACPI]
25 { vendor | video | native | none }
26 If set to vendor, prefer vendor-specific driver
27 (e.g. thinkpad_acpi, sony_acpi, etc.) instead
28 of the ACPI video.ko driver.
29 If set to video, use the ACPI video.ko driver.
30 If set to native, use the device's native backlight mode.
31 If set to none, disable the ACPI backlight interface.
33 acpi_force_32bit_fadt_addr
34 force FADT to use 32 bit addresses rather than the
35 64 bit X_* addresses. Some firmware have broken 64
36 bit addresses for force ACPI ignore these and use
37 the older legacy 32 bit addresses.
39 acpica_no_return_repair [HW, ACPI]
40 Disable AML predefined validation mechanism
41 This mechanism can repair the evaluation result to make
42 the return objects more ACPI specification compliant.
43 This option is useful for developers to identify the
44 root cause of an AML interpreter issue when the issue
45 has something to do with the repair mechanism.
47 acpi.debug_layer= [HW,ACPI,ACPI_DEBUG]
48 acpi.debug_level= [HW,ACPI,ACPI_DEBUG]
50 CONFIG_ACPI_DEBUG must be enabled to produce any ACPI
51 debug output. Bits in debug_layer correspond to a
52 _COMPONENT in an ACPI source file, e.g.,
53 #define _COMPONENT ACPI_EVENTS
54 Bits in debug_level correspond to a level in
55 ACPI_DEBUG_PRINT statements, e.g.,
56 ACPI_DEBUG_PRINT((ACPI_DB_INFO, ...
57 The debug_level mask defaults to "info". See
58 Documentation/firmware-guide/acpi/debug.rst for more information about
59 debug layers and levels.
61 Enable processor driver info messages:
62 acpi.debug_layer=0x20000000
63 Enable AML "Debug" output, i.e., stores to the Debug
64 object while interpreting AML:
65 acpi.debug_layer=0xffffffff acpi.debug_level=0x2
66 Enable all messages related to ACPI hardware:
67 acpi.debug_layer=0x2 acpi.debug_level=0xffffffff
69 Some values produce so much output that the system is
70 unusable. The "log_buf_len" parameter may be useful
71 if you need to capture more output.
73 acpi_enforce_resources= [ACPI]
75 Check for resource conflicts between native drivers
76 and ACPI OperationRegions (SystemIO and SystemMemory
77 only). IO ports and memory declared in ACPI might be
78 used by the ACPI subsystem in arbitrary AML code and
79 can interfere with legacy drivers.
80 strict (default): access to resources claimed by ACPI
81 is denied; legacy drivers trying to access reserved
82 resources will fail to bind to device using them.
83 lax: access to resources claimed by ACPI is allowed;
84 legacy drivers trying to access reserved resources
85 will bind successfully but a warning message is logged.
86 no: ACPI OperationRegions are not marked as reserved,
87 no further checks are performed.
89 acpi_force_table_verification [HW,ACPI]
90 Enable table checksum verification during early stage.
91 By default, this is disabled due to x86 early mapping
94 acpi_irq_balance [HW,ACPI]
95 ACPI will balance active IRQs
98 acpi_irq_nobalance [HW,ACPI]
99 ACPI will not move active IRQs (default)
102 acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA
103 Format: <irq>,<irq>...
105 acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for
107 Format: <irq>,<irq>...
109 acpi_mask_gpe= [HW,ACPI]
110 Due to the existence of _Lxx/_Exx, some GPEs triggered
111 by unsupported hardware/firmware features can result in
112 GPE floodings that cannot be automatically disabled by
114 This facility can be used to prevent such uncontrolled
116 Format: <byte> or <bitmap-list>
118 acpi_no_auto_serialize [HW,ACPI]
119 Disable auto-serialization of AML methods
120 AML control methods that contain the opcodes to create
121 named objects will be marked as "Serialized" by the
122 auto-serialization feature.
123 This feature is enabled by default.
124 This option allows to turn off the feature.
126 acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump
129 acpi_no_static_ssdt [HW,ACPI]
130 Disable installation of static SSDTs at early boot time
131 By default, SSDTs contained in the RSDT/XSDT will be
132 installed automatically and they will appear under
133 /sys/firmware/acpi/tables.
134 This option turns off this feature.
135 Note that specifying this option does not affect
136 dynamic table installation which will install SSDT
137 tables to /sys/firmware/acpi/tables/dynamic.
139 acpi_no_watchdog [HW,ACPI,WDT]
140 Ignore the ACPI-based watchdog interface (WDAT) and let
141 a native driver control the watchdog device instead.
143 acpi_rsdp= [ACPI,EFI,KEXEC]
144 Pass the RSDP address to the kernel, mostly used
145 on machines running EFI runtime service to boot the
146 second kernel for kdump.
148 acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS
149 Format: To spoof as Windows 98: ="Microsoft Windows"
151 acpi_rev_override [ACPI] Override the _REV object to return 5 (instead
152 of 2 which is mandated by ACPI 6) as the supported ACPI
153 specification revision (when using this switch, it may
154 be necessary to carry out a cold reboot _twice_ in a
155 row to make it take effect on the platform firmware).
157 acpi_osi= [HW,ACPI] Modify list of supported OS interface strings
158 acpi_osi="string1" # add string1
159 acpi_osi="!string2" # remove string2
160 acpi_osi=!* # remove all strings
161 acpi_osi=! # disable all built-in OS vendor
163 acpi_osi=!! # enable all built-in OS vendor
165 acpi_osi= # disable all strings
167 'acpi_osi=!' can be used in combination with single or
168 multiple 'acpi_osi="string1"' to support specific OS
169 vendor string(s). Note that such command can only
170 affect the default state of the OS vendor strings, thus
171 it cannot affect the default state of the feature group
172 strings and the current state of the OS vendor strings,
173 specifying it multiple times through kernel command line
174 is meaningless. This command is useful when one do not
175 care about the state of the feature group strings which
176 should be controlled by the OSPM.
178 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent
179 to 'acpi_osi="Windows 2000" acpi_osi=!', they all
180 can make '_OSI("Windows 2000")' TRUE.
182 'acpi_osi=' cannot be used in combination with other
183 'acpi_osi=' command lines, the _OSI method will not
184 exist in the ACPI namespace. NOTE that such command can
185 only affect the _OSI support state, thus specifying it
186 multiple times through kernel command line is also
189 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)'
192 'acpi_osi=!*' can be used in combination with single or
193 multiple 'acpi_osi="string1"' to support specific
194 string(s). Note that such command can affect the
195 current state of both the OS vendor strings and the
196 feature group strings, thus specifying it multiple times
197 through kernel command line is meaningful. But it may
198 still not able to affect the final state of a string if
199 there are quirks related to this string. This command
200 is useful when one want to control the state of the
201 feature group strings to debug BIOS issues related to
204 1. 'acpi_osi="Module Device" acpi_osi=!*' can make
205 '_OSI("Module Device")' FALSE.
206 2. 'acpi_osi=!* acpi_osi="Module Device"' can make
207 '_OSI("Module Device")' TRUE.
208 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is
210 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"'
212 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!',
213 they all will make '_OSI("Windows 2000")' TRUE.
216 Override the pmtimer bug detection: force the kernel
217 to assume that this machine's pmtimer latches its value
218 and always returns good values.
220 acpi_sci= [HW,ACPI] ACPI System Control Interrupt trigger mode
221 Format: { level | edge | high | low }
223 acpi_skip_timer_override [HW,ACPI]
224 Recognize and ignore IRQ0/pin2 Interrupt Override.
225 For broken nForce2 BIOS resulting in XT-PIC timer.
227 acpi_sleep= [HW,ACPI] Sleep options
228 Format: { s3_bios, s3_mode, s3_beep, s4_hwsig,
229 s4_nohwsig, old_ordering, nonvs,
230 sci_force_enable, nobl }
231 See Documentation/power/video.rst for information on
233 s3_beep is for debugging; it makes the PC's speaker beep
234 as soon as the kernel's real-mode entry point is called.
235 s4_hwsig causes the kernel to check the ACPI hardware
236 signature during resume from hibernation, and gracefully
237 refuse to resume if it has changed. This complies with
238 the ACPI specification but not with reality, since
239 Windows does not do this and many laptops do change it
240 on docking. So the default behaviour is to allow resume
241 and simply warn when the signature changes, unless the
242 s4_hwsig option is enabled.
243 s4_nohwsig prevents ACPI hardware signature from being
244 used (or even warned about) during resume.
245 old_ordering causes the ACPI 1.0 ordering of the _PTS
246 control method, with respect to putting devices into
247 low power states, to be enforced (the ACPI 2.0 ordering
248 of _PTS is used by default).
249 nonvs prevents the kernel from saving/restoring the
250 ACPI NVS memory during suspend/hibernation and resume.
251 sci_force_enable causes the kernel to set SCI_EN directly
252 on resume from S1/S3 (which is against the ACPI spec,
253 but some broken systems don't work without it).
254 nobl causes the internal blacklist of systems known to
255 behave incorrectly in some ways with respect to system
256 suspend and resume to be ignored (use wisely).
258 acpi_use_timer_override [HW,ACPI]
259 Use timer override. For some broken Nvidia NF5 boards
260 that require a timer override, but don't have HPET
262 add_efi_memmap [EFI; X86] Include EFI memory map in
263 kernel's map of available physical RAM.
266 { off | try_unsupported }
267 off: disable AGP support
268 try_unsupported: try to drive unsupported chipsets
269 (may crash computer or cause data corruption)
272 See Documentation/sound/alsa-configuration.rst
275 Allow the default userspace alignment fault handler
276 behaviour to be specified. Bit 0 enables warnings,
277 bit 1 enables fixups, and bit 2 sends a segfault.
279 align_va_addr= [X86-64]
280 Align virtual addresses by clearing slice [14:12] when
281 allocating a VMA at process creation time. This option
282 gives you up to 3% performance improvement on AMD F15h
283 machines (where it is enabled by default) for a
284 CPU-intensive style benchmark, and it can vary highly in
285 a microbenchmark depending on workload and compiler.
287 32: only for 32-bit processes
288 64: only for 64-bit processes
289 on: enable for both 32- and 64-bit processes
290 off: disable for both 32- and 64-bit processes
292 alloc_snapshot [FTRACE]
293 Allocate the ftrace snapshot buffer on boot up when the
294 main buffer is allocated. This is handy if debugging
295 and you need to use tracing_snapshot() on boot up, and
296 do not want to use tracing_snapshot_alloc() as it needs
297 to be done where GFP_KERNEL allocations are allowed.
299 allow_mismatched_32bit_el0 [ARM64]
300 Allow execve() of 32-bit applications and setting of the
301 PER_LINUX32 personality on systems where only a strict
302 subset of the CPUs support 32-bit EL0. When this
303 parameter is present, the set of CPUs supporting 32-bit
304 EL0 is indicated by /sys/devices/system/cpu/aarch32_el0
305 and hot-unplug operations may be restricted.
307 See Documentation/arm64/asymmetric-32bit.rst for more
310 amd_iommu= [HW,X86-64]
311 Pass parameters to the AMD IOMMU driver in the system.
313 fullflush - Deprecated, equivalent to iommu.strict=1
314 off - do not initialize any AMD IOMMU found in
316 force_isolation - Force device isolation for all
317 devices. The IOMMU driver is not
318 allowed anymore to lift isolation
319 requirements as needed. This option
320 does not override iommu=pt
321 force_enable - Force enable the IOMMU on platforms known
322 to be buggy with IOMMU enabled. Use this
325 amd_iommu_dump= [HW,X86-64]
326 Enable AMD IOMMU driver option to dump the ACPI table
327 for AMD IOMMU. With this option enabled, AMD IOMMU
328 driver will print ACPI tables for AMD IOMMU during
329 IOMMU initialization.
331 amd_iommu_intr= [HW,X86-64]
332 Specifies one of the following AMD IOMMU interrupt
334 legacy - Use legacy interrupt remapping mode.
335 vapic - Use virtual APIC mode, which allows IOMMU
336 to inject interrupts directly into guest.
337 This mode requires kvm-amd.avic=1.
338 (Default when IOMMU HW support is present.)
340 amijoy.map= [HW,JOY] Amiga joystick support
341 Map of devices attached to JOY0DAT and JOY1DAT
343 See also Documentation/input/joydev/joystick.rst
345 analog.map= [HW,JOY] Analog joystick and gamepad support
346 Specifies type or capabilities of an analog joystick
347 connected to one of 16 gameports
348 Format: <type1>,<type2>,..<type16>
351 Power management functions (SPARCstation-4/5 + deriv.)
353 Disable APC CPU standby support. SPARCstation-Fox does
354 not play well with APC CPU idle - disable it if you have
355 APC and your system crashes randomly.
357 apic= [APIC,X86] Advanced Programmable Interrupt Controller
358 Change the output verbosity while booting
359 Format: { quiet (default) | verbose | debug }
360 Change the amount of debugging information output
361 when initialising the APIC and IO-APIC components.
362 For X86-32, this can also be used to specify an APIC
364 Format: apic=driver_name
365 Examples: apic=bigsmp
367 apic_extnmi= [APIC,X86] External NMI delivery setting
368 Format: { bsp (default) | all | none }
369 bsp: External NMI is delivered only to CPU 0
370 all: External NMIs are broadcast to all CPUs as a
372 none: External NMI is masked for all CPUs. This is
373 useful so that a dump capture kernel won't be
377 See Documentation/networking/ipv6.rst.
379 show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller
380 Limit apic dumping. The parameter defines the maximal
381 number of local apics being dumped. Also it is possible
382 to set it to "all" by meaning -- no limit here.
383 Format: { 1 (default) | 2 | ... | all }.
384 The parameter valid if only apic=debug or
385 apic=verbose is specified.
386 Example: apic=debug show_lapic=all
388 apm= [APM] Advanced Power Management
389 See header of arch/x86/kernel/apm_32.c.
391 arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards
392 Format: <io>,<irq>,<nodeID>
394 arm64.nobti [ARM64] Unconditionally disable Branch Target
395 Identification support
397 arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication
400 arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension
405 atarimouse= [HW,MOUSE] Atari Mouse
407 atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess,
408 EzKey and similar keyboards
410 atkbd.reset= [HW] Reset keyboard during initialization
412 atkbd.set= [HW] Select keyboard code set
413 Format: <int> (2 = AT (default), 3 = PS/2)
415 atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar
418 atkbd.softraw= [HW] Choose between synthetic and real raw mode
419 Format: <bool> (0 = real, 1 = synthetic (default))
421 atkbd.softrepeat= [HW]
422 Use software keyboard repeat
424 audit= [KNL] Enable the audit sub-system
425 Format: { "0" | "1" | "off" | "on" }
426 0 | off - kernel audit is disabled and can not be
427 enabled until the next reboot
428 unset - kernel audit is initialized but disabled and
429 will be fully enabled by the userspace auditd.
430 1 | on - kernel audit is initialized and partially
431 enabled, storing at most audit_backlog_limit
432 messages in RAM until it is fully enabled by the
436 audit_backlog_limit= [KNL] Set the audit queue size limit.
437 Format: <int> (must be >=0)
440 bau= [X86_UV] Enable the BAU on SGI UV. The default
441 behavior is to disable the BAU (i.e. bau=0).
442 Format: { "0" | "1" }
445 unset - Disable the BAU.
447 baycom_epp= [HW,AX25]
450 baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem
452 See header of drivers/net/hamradio/baycom_par.c.
454 baycom_ser_fdx= [HW,AX25]
455 BayCom Serial Port AX.25 Modem (Full Duplex Mode)
456 Format: <io>,<irq>,<mode>[,<baud>]
457 See header of drivers/net/hamradio/baycom_ser_fdx.c.
459 baycom_ser_hdx= [HW,AX25]
460 BayCom Serial Port AX.25 Modem (Half Duplex Mode)
461 Format: <io>,<irq>,<mode>
462 See header of drivers/net/hamradio/baycom_ser_hdx.c.
465 Disable BERT OS support on buggy BIOSes.
467 bgrt_disable [ACPI][X86]
468 Disable BGRT to avoid flickering OEM logo.
470 blkdevparts= Manual partition parsing of block device(s) for
471 embedded devices based on command line input.
472 See Documentation/block/cmdline-partition.rst
474 boot_delay= Milliseconds to delay each printk during boot.
475 Values larger than 10 seconds (10000) are changed to
480 Extended command line options can be added to an initrd
481 and this will cause the kernel to look for it.
483 See Documentation/admin-guide/bootconfig.rst
485 bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards)
486 bttv.radio= Most important insmod options are available as
488 bttv.pll= See Documentation/admin-guide/media/bttv.rst
491 bulk_remove=off [PPC] This parameter disables the use of the pSeries
492 firmware feature for flushing multiple hpte entries
495 c101= [NET] Moxa C101 synchronous serial card
497 cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection.
498 Sometimes CPU hardware bugs make them report the cache
499 size incorrectly. The kernel will attempt work arounds
500 to fix known problems, but for some CPUs it is not
501 possible to determine what the correct size should be.
502 This option provides an override for these situations.
505 [NET] Specifies amount of time (in seconds) that
506 the kernel should wait for a network carrier. By default
507 it waits 120 seconds.
509 ca_keys= [KEYS] This parameter identifies a specific key(s) on
510 the system trusted keyring to be used for certificate
512 format: { id:<keyid> | builtin }
514 cca= [MIPS] Override the kernel pages' cache coherency
515 algorithm. Accepted values range from 0 to 7
516 inclusive. See arch/mips/include/asm/pgtable-bits.h
517 for platform specific values (SB1, Loongson3 and
520 ccw_timeout_log [S390]
521 See Documentation/s390/common_io.rst for details.
523 cgroup_disable= [KNL] Disable a particular controller or optional feature
524 Format: {name of the controller(s) or feature(s) to disable}
525 The effects of cgroup_disable=foo are:
526 - foo isn't auto-mounted if you mount all cgroups in
528 - foo isn't visible as an individually mountable
530 - if foo is an optional feature then the feature is
531 disabled and corresponding cgroup files are not
533 {Currently only "memory" controller deal with this and
534 cut the overhead, others just disable the usage. So
535 only cgroup_disable=memory is actually worthy}
536 Specifying "pressure" disables per-cgroup pressure
537 stall information accounting feature
539 cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1
540 Format: { { controller | "all" | "named" }
541 [,{ controller | "all" | "named" }...] }
542 Like cgroup_disable, but only applies to cgroup v1;
543 the blacklisted controllers remain available in cgroup2.
544 "all" blacklists all controllers and "named" disables
545 named mounts. Specifying both "all" and "named" disables
548 cgroup.memory= [KNL] Pass options to the cgroup memory controller.
550 nosocket -- Disable socket memory accounting.
551 nokmem -- Disable kernel memory accounting.
553 checkreqprot [SELINUX] Set initial checkreqprot flag value.
554 Format: { "0" | "1" }
555 See security/selinux/Kconfig help text.
556 0 -- check protection applied by kernel (includes
557 any implied execute protection).
558 1 -- check protection requested by application.
559 Default value is set via a kernel config option.
560 Value can be changed at runtime via
561 /sys/fs/selinux/checkreqprot.
562 Setting checkreqprot to 1 is deprecated.
565 See Documentation/s390/common_io.rst for details.
567 clearcpuid=X[,X...] [X86]
568 Disable CPUID feature X for the kernel. See
569 arch/x86/include/asm/cpufeatures.h for the valid bit
570 numbers X. Note the Linux-specific bits are not necessarily
571 stable over kernel options, but the vendor-specific
573 X can also be a string as appearing in the flags: line
574 in /proc/cpuinfo which does not have the above
575 instability issue. However, not all features have names
577 Note that using this option will taint your kernel.
578 Also note that user programs calling CPUID directly
579 or using the feature without checking anything
580 will still see it. This just prevents it from
581 being used by the kernel or shown in /proc/cpuinfo.
582 Also note the kernel might malfunction if you disable
587 Prevents the clock framework from automatically gating
588 clocks that have not been explicitly enabled by a Linux
589 device driver but are enabled in hardware at reset or
590 by the bootloader/firmware. Note that this does not
591 force such clocks to be always-on nor does it reserve
592 those clocks in any way. This parameter is useful for
593 debug and development, but should not be needed on a
594 platform with proper driver support. For more
595 information, see Documentation/driver-api/clk.rst.
597 clock= [BUGS=X86-32, HW] gettimeofday clocksource override.
599 Forces specified clocksource (if available) to be used
600 when calculating gettimeofday(). If specified
601 clocksource is not available, it defaults to PIT.
602 Format: { pit | tsc | cyclone | pmtmr }
604 clocksource= Override the default clocksource
606 Override the default clocksource and use the clocksource
607 with the name specified.
608 Some clocksource names to choose from, depending on
610 [all] jiffies (this is the base, fallback clocksource)
612 [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
613 pxa_timer,timer3,32k_counter,timer0_1
614 [X86-32] pit,hpet,tsc;
615 scx200_hrt on Geode; cyclone on IBM x440
623 clocksource.arm_arch_timer.evtstrm=
626 Enable/disable the eventstream feature of the ARM
627 architected timer so that code using WFE-based polling
628 loops can be debugged more effectively on production
631 clocksource.max_cswd_read_retries= [KNL]
632 Number of clocksource_watchdog() retries due to
633 external delays before the clock will be marked
634 unstable. Defaults to two retries, that is,
635 three attempts to read the clock under test.
637 clocksource.verify_n_cpus= [KNL]
638 Limit the number of CPUs checked for clocksources
639 marked with CLOCK_SOURCE_VERIFY_PERCPU that
640 are marked unstable due to excessive skew.
641 A negative value says to check all CPUs, while
642 zero says not to check any. Values larger than
643 nr_cpu_ids are silently truncated to nr_cpu_ids.
644 The actual CPUs are chosen randomly, with
645 no replacement if the same CPU is chosen twice.
647 clocksource-wdtest.holdoff= [KNL]
648 Set the time in seconds that the clocksource
649 watchdog test waits before commencing its tests.
650 Defaults to zero when built as a module and to
651 10 seconds when built into the kernel.
653 cma=nn[MG]@[start[MG][-end[MG]]]
655 Sets the size of kernel global memory area for
656 contiguous memory allocations and optionally the
657 placement constraint by the physical address range of
658 memory allocations. A value of 0 disables CMA
659 altogether. For more information, see
660 kernel/dma/contiguous.c
664 Sets the size of kernel per-numa memory area for
665 contiguous memory allocations. A value of 0 disables
666 per-numa CMA altogether. And If this option is not
667 specificed, the default value is 0.
668 With per-numa CMA enabled, DMA users on node nid will
669 first try to allocate buffer from the pernuma area
670 which is located in node nid, if the allocation fails,
671 they will fallback to the global default memory area.
673 cmo_free_hint= [PPC] Format: { yes | no }
674 Specify whether pages are marked as being inactive
675 when they are freed. This is used in CMO environments
676 to determine OS memory pressure for page stealing by
680 coherent_pool=nn[KMG] [ARM,KNL]
681 Sets the size of memory pool for coherent, atomic dma
682 allocations, by default set to 256K.
684 com20020= [HW,NET] ARCnet - COM20020 chipset
686 <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]]
688 com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers)
692 ARCnet - COM90xx chipset (memory-mapped buffers)
693 Format: <io>[,<irq>[,<memstart>]]
695 condev= [HW,S390] console device
698 console= [KNL] Output console device and options.
700 tty<n> Use the virtual console device <n>.
704 Use the specified serial port. The options are of
705 the form "bbbbpnf", where "bbbb" is the baud rate,
706 "p" is parity ("n", "o", or "e"), "n" is number of
707 bits, and "f" is flow control ("r" for RTS or
708 omit it). Default is "9600n8".
710 See Documentation/admin-guide/serial-console.rst for more
712 Documentation/networking/netconsole.rst for an
715 uart[8250],io,<addr>[,options]
716 uart[8250],mmio,<addr>[,options]
717 uart[8250],mmio16,<addr>[,options]
718 uart[8250],mmio32,<addr>[,options]
719 uart[8250],0x<addr>[,options]
720 Start an early, polled-mode console on the 8250/16550
721 UART at the specified I/O port or MMIO address,
722 switching to the matching ttyS device later.
723 MMIO inter-register address stride is either 8-bit
724 (mmio), 16-bit (mmio16), or 32-bit (mmio32).
725 If none of [io|mmio|mmio16|mmio32], <addr> is assumed
726 to be equivalent to 'mmio'. 'options' are specified in
727 the same format described for ttyS above; if unspecified,
728 the h/w is not re-initialized.
730 hvc<n> Use the hypervisor console device <n>. This is for
731 both Xen and PowerPC hypervisors.
734 Use to disable console output, i.e., to have kernel
735 console messages discarded.
736 This must be the only console= parameter used on the
739 If the device connected to the port is not a TTY but a braille
740 device, prepend "brl," before the device type, for instance
742 For now, only VisioBraille is supported.
745 [KNL] Change console messages format
747 By default we print messages on consoles in
748 "[time stamp] text\n" format (time stamp may not be
749 printed, depending on CONFIG_PRINTK_TIME or
750 `printk_time' param).
752 Switch to syslog format: "<%u>[time stamp] text\n"
753 IOW, each message will have a facility and loglevel
754 prefix. The format is similar to one used by syslog()
755 syscall, or to executing "dmesg -S --raw" or to reading
758 consoleblank= [KNL] The console blank (screen saver) timeout in
759 seconds. A value of 0 disables the blank timer.
763 [KNL] Change the default value for
764 /proc/<pid>/coredump_filter.
765 See also Documentation/filesystems/proc.rst.
767 coresight_cpu_debug.enable
770 Enable/disable the CPU sampling based debugging.
771 0: default value, disable debugging
772 1: enable debugging at boot time
774 cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver
776 <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>]
778 cpu0_hotplug [X86] Turn on CPU0 hotplug feature when
779 CONFIG_BOOTPARAM_HOTPLUG_CPU0 is off.
780 Some features depend on CPU0. Known dependencies are:
781 1. Resume from suspend/hibernate depends on CPU0.
782 Suspend/hibernate will fail if CPU0 is offline and you
783 need to online CPU0 before suspend/hibernate.
784 2. PIC interrupts also depend on CPU0. CPU0 can't be
785 removed if a PIC interrupt is detected.
786 It's said poweroff/reboot may depend on CPU0 on some
787 machines although I haven't seen such issues so far
788 after CPU0 is offline on a few tested machines.
789 If the dependencies are under your control, you can
790 turn on cpu0_hotplug.
792 cpuidle.off=1 [CPU_IDLE]
793 disable the cpuidle sub-system
796 [CPU_IDLE] Name of the cpuidle governor to use.
798 cpufreq.off=1 [CPU_FREQ]
799 disable the cpufreq sub-system
801 cpufreq.default_governor=
802 [CPU_FREQ] Name of the default cpufreq governor or
803 policy to use. This governor must be registered in the
804 kernel before the cpufreq driver probes.
807 [X86] Delay for N microsec between assert and de-assert
808 of APIC INIT to start processors. This delay occurs
809 on every CPU online, such as boot, and resume from suspend.
812 crash_kexec_post_notifiers
813 Run kdump after running panic-notifiers and dumping
814 kmsg. This only for the users who doubt kdump always
815 succeeds in any situation.
816 Note that this also increases risks of kdump failure,
817 because some panic notifiers can make the crashed
818 kernel more unstable.
820 crashkernel=size[KMG][@offset[KMG]]
821 [KNL] Using kexec, Linux can switch to a 'crash kernel'
822 upon panic. This parameter reserves the physical
823 memory region [offset, offset + size] for that kernel
824 image. If '@offset' is omitted, then a suitable offset
825 is selected automatically.
826 [KNL, X86-64] Select a region under 4G first, and
827 fall back to reserve region above 4G when '@offset'
828 hasn't been specified.
829 See Documentation/admin-guide/kdump/kdump.rst for further details.
831 crashkernel=range1:size1[,range2:size2,...][@offset]
832 [KNL] Same as above, but depends on the memory
833 in the running system. The syntax of range is
834 start-[end] where start and end are both
835 a memory unit (amount[KMG]). See also
836 Documentation/admin-guide/kdump/kdump.rst for an example.
838 crashkernel=size[KMG],high
839 [KNL, X86-64, ARM64] range could be above 4G. Allow kernel
840 to allocate physical memory region from top, so could
841 be above 4G if system have more than 4G ram installed.
842 Otherwise memory region will be allocated below 4G, if
844 It will be ignored if crashkernel=X is specified.
845 crashkernel=size[KMG],low
846 [KNL, X86-64] range under 4G. When crashkernel=X,high
847 is passed, kernel could allocate physical memory region
848 above 4G, that cause second kernel crash on system
849 that require some amount of low memory, e.g. swiotlb
850 requires at least 64M+32K low memory, also enough extra
851 low memory is needed to make sure DMA buffers for 32-bit
852 devices won't run out. Kernel would try to allocate
853 at least 256M below 4G automatically.
854 This one lets the user specify own low range under 4G
855 for second kernel instead.
856 0: to disable low allocation.
857 It will be ignored when crashkernel=X,high is not used
858 or memory reserved is below 4G.
860 [KNL, ARM64] range in low memory.
861 This one lets the user specify a low range in the
862 DMA zone for the crash dump kernel.
863 It will be ignored when crashkernel=X,high is not used
864 or memory reserved is located in the DMA zones.
867 [KNL] Disable crypto self-tests
872 cs89x0_media= [HW,NET]
873 Format: { rj45 | aui | bnc }
875 csdlock_debug= [KNL] Enable debug add-ons of cross-CPU function call
876 handling. When switched on, additional debug data is
877 printed to the console in case a hanging CPU is
878 detected, and that CPU is pinged again in order to try
879 to resolve the hang situation.
880 0: disable csdlock debugging (default)
881 1: enable basic csdlock debugging (minor impact)
882 ext: enable extended csdlock debugging (more impact,
886 See header of drivers/s390/block/dasd_devmap.c.
888 db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port
889 (one device per port)
890 Format: <port#>,<type>
891 See also Documentation/input/devices/joystick-parport.rst
893 debug [KNL] Enable kernel debugging (events log level).
896 [KNL] Enable printing [hashed] pointers early in the
897 boot sequence. If enabled, we use a weak hash instead
898 of siphash to hash pointers. Use this option if you are
899 seeing instances of '(___ptrval___)') and need to see a
900 value (hashed pointer) instead. Cryptographically
901 insecure, please do not use on production kernels.
904 [KNL] verbose locking self-tests
906 Print debugging info while doing the locking API
908 Bitmask for the various LOCKTYPE_ tests. Defaults to 0
909 (no extra messages), setting it to -1 (all bits set)
910 will print _a_lot_ more information - normally only
911 useful to lockdep developers.
913 debug_objects [KNL] Enable object debugging
916 [KNL] Disable object debugging
918 debug_guardpage_minorder=
919 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this
920 parameter allows control of the order of pages that will
921 be intentionally kept free (and hence protected) by the
922 buddy allocator. Bigger value increase the probability
923 of catching random memory corruption, but reduce the
924 amount of memory for normal system use. The maximum
925 possible value is MAX_ORDER/2. Setting this parameter
926 to 1 or 2 should be enough to identify most random
927 memory corruption problems caused by bugs in kernel or
928 driver code when a CPU writes to (or reads from) a
929 random memory location. Note that there exists a class
930 of memory corruptions problems caused by buggy H/W or
931 F/W or by drivers badly programing DMA (basically when
932 memory is written at bus level and the CPU MMU is
933 bypassed) which are not detectable by
934 CONFIG_DEBUG_PAGEALLOC, hence this option will not help
935 tracking down these problems.
938 [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter
939 enables the feature at boot time. By default, it is
940 disabled and the system will work mostly the same as a
941 kernel built without CONFIG_DEBUG_PAGEALLOC.
942 Note: to get most of debug_pagealloc error reports, it's
943 useful to also enable the page_owner functionality.
944 on: enable the feature
946 debugfs= [KNL] This parameter enables what is exposed to userspace
947 and debugfs internal clients.
948 Format: { on, no-mount, off }
949 on: All functions are enabled.
951 Filesystem is not registered but kernel clients can
952 access APIs and a crashkernel can be used to read
953 its content. There is nothing to mount.
954 off: Filesystem is not registered and clients
955 get a -EPERM as result when trying to register files
956 or directories within debugfs.
957 This is equivalent of the runtime functionality if
958 debugfs was not enabled in the kernel at all.
959 Default value is set in build-time with a kernel configuration.
961 debugpat [X86] Enable PAT debugging
963 decnet.addr= [HW,NET]
964 Format: <area>[,<node>]
965 See also Documentation/networking/decnet.rst.
968 [HW] The size of the default HugeTLB page. This is
969 the size represented by the legacy /proc/ hugepages
970 APIs. In addition, this is the default hugetlb size
971 used for shmget(), mmap() and mounting hugetlbfs
972 filesystems. If not specified, defaults to the
973 architecture's default huge page size. Huge page
974 sizes are architecture dependent. See also
975 Documentation/admin-guide/mm/hugetlbpage.rst.
978 deferred_probe_timeout=
979 [KNL] Debugging option to set a timeout in seconds for
980 deferred probe to give up waiting on dependencies to
981 probe. Only specific dependencies (subsystems or
982 drivers) that have opted in will be ignored. A timeout of 0
983 will timeout at the end of initcalls. This option will also
984 dump out devices still on the deferred probe list after
987 delayacct [KNL] Enable per-task delay accounting
989 dell_smm_hwmon.ignore_dmi=
990 [HW] Continue probing hardware even if DMI data
991 indicates that the driver is running on unsupported
994 dell_smm_hwmon.force=
995 [HW] Activate driver even if SMM BIOS signature does
996 not match list of supported models and enable otherwise
997 blacklisted features.
999 dell_smm_hwmon.power_status=
1000 [HW] Report power status in /proc/i8k
1001 (disabled by default).
1003 dell_smm_hwmon.restricted=
1004 [HW] Allow controlling fans only if SYS_ADMIN
1007 dell_smm_hwmon.fan_mult=
1008 [HW] Factor to multiply fan speed with.
1010 dell_smm_hwmon.fan_max=
1011 [HW] Maximum configurable fan speed.
1014 Format: { on | off | def_only | inf_only | always }
1015 on: s390 zlib hardware support for compression on
1016 level 1 and decompression (default)
1017 off: No s390 zlib hardware support
1018 def_only: s390 zlib hardware support for deflate
1019 only (compression on level 1)
1020 inf_only: s390 zlib hardware support for inflate
1021 only (decompression)
1022 always: Same as 'on' but ignores the selected compression
1023 level always using hardware support (used for debugging)
1025 dhash_entries= [KNL]
1026 Set number of hash buckets for dentry cache.
1028 disable_1tb_segments [PPC]
1029 Disables the use of 1TB hash page table segments. This
1030 causes the kernel to fall back to 256MB segments which
1031 can be useful when debugging issues that require an SLB
1035 Limits the number of kernel SLB entries, and flushes
1036 them frequently to increase the rate of SLB faults
1037 on kernel addresses.
1040 See Documentation/networking/ipv6.rst.
1043 Disable RADIX MMU mode on POWER9
1045 radix_hcall_invalidate=on [PPC/PSERIES]
1046 Disable RADIX GTSE feature and use hcall for TLB
1050 Disable TLBIE instruction. Currently does not work
1051 with KVM, with HASH MMU, or with coherent accelerators.
1053 disable_cpu_apicid= [X86,APIC,SMP]
1055 The number of initial APIC ID for the
1056 corresponding CPU to be disabled at boot,
1057 mostly used for the kdump 2nd kernel to
1058 disable BSP to wake up multiple CPUs without
1059 causing system reset or hang due to sending
1060 INIT from AP to BSP.
1062 disable_ddw [PPC/PSERIES]
1063 Disable Dynamic DMA Window support. Use this
1064 to workaround buggy firmware.
1066 disable_ipv6= [IPV6]
1067 See Documentation/networking/ipv6.rst.
1069 disable_mtrr_cleanup [X86]
1070 The kernel tries to adjust MTRR layout from continuous
1071 to discrete, to make X server driver able to add WB
1072 entry later. This parameter disables that.
1074 disable_mtrr_trim [X86, Intel and AMD only]
1075 By default the kernel will trim any uncacheable
1076 memory out of your available memory pool based on
1077 MTRR settings. This parameter disables that behavior,
1078 possibly causing your machine to run very slowly.
1080 disable_timer_pin_1 [X86]
1081 Disable PIN 1 of APIC timer
1082 Can be useful to work around chipset bugs.
1084 dis_ucode_ldr [X86] Disable the microcode loader.
1086 dma_debug=off If the kernel is compiled with DMA_API_DEBUG support,
1087 this option disables the debugging code at boot.
1089 dma_debug_entries=<number>
1090 This option allows to tune the number of preallocated
1091 entries for DMA-API debugging code. One entry is
1092 required per DMA-API allocation. Use this if the
1093 DMA-API debugging code disables itself because the
1094 architectural default is too low.
1096 dma_debug_driver=<driver_name>
1097 With this option the DMA-API debugging driver
1098 filter feature can be enabled at boot time. Just
1099 pass the driver to filter for as the parameter.
1100 The filter can be disabled or changed to another
1101 driver later using sysfs.
1103 driver_async_probe= [KNL]
1104 List of driver names to be probed asynchronously.
1105 Format: <driver_name1>,<driver_name2>...
1107 drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
1108 Broken monitors, graphic adapters, KVMs and EDIDless
1109 panels may send no or incorrect EDID data sets.
1110 This parameter allows to specify an EDID data sets
1111 in the /lib/firmware directory that are used instead.
1112 Generic built-in EDID data sets are used, if one of
1113 edid/1024x768.bin, edid/1280x1024.bin,
1114 edid/1680x1050.bin, or edid/1920x1080.bin is given
1115 and no file with the same name exists. Details and
1116 instructions how to build your own EDID data are
1117 available in Documentation/admin-guide/edid.rst. An EDID
1118 data set will only be used for a particular connector,
1119 if its name and a colon are prepended to the EDID
1120 name. Each connector may use a unique EDID data
1121 set by separating the files with a comma. An EDID
1122 data set with no connector name will be used for
1123 any connectors not explicitly specified.
1128 Format: {"off" | "known"}
1129 Control how the dt_cpu_ftrs device-tree binding is
1130 used for CPU feature discovery and setup (if it
1132 off: Do not use it, fall back to legacy cpu table.
1133 known: Do not pass through unknown features to guests
1134 or userspace, only those that the kernel is aware of.
1136 dump_apple_properties [X86]
1137 Dump name and content of EFI device properties on
1138 x86 Macs. Useful for driver authors to determine
1139 what data is available or for reverse-engineering.
1141 dyndbg[="val"] [KNL,DYNAMIC_DEBUG]
1142 <module>.dyndbg[="val"]
1143 Enable debug messages at boot time. See
1144 Documentation/admin-guide/dynamic-debug-howto.rst
1147 nopku [X86] Disable Memory Protection Keys CPU feature found
1150 <module>.async_probe [KNL]
1151 Enable asynchronous probe on this module.
1153 early_ioremap_debug [KNL]
1154 Enable debug messages in early_ioremap support. This
1155 is useful for tracking down temporary early mappings
1156 which are not unmapped.
1158 earlycon= [KNL] Output early console device and options.
1160 When used with no options, the early console is
1161 determined by stdout-path property in device tree's
1162 chosen node or the ACPI SPCR table if supported by
1165 cdns,<addr>[,options]
1166 Start an early, polled-mode console on a Cadence
1167 (xuartps) serial port at the specified address. Only
1168 supported option is baud rate. If baud rate is not
1169 specified, the serial port must already be setup and
1172 uart[8250],io,<addr>[,options]
1173 uart[8250],mmio,<addr>[,options]
1174 uart[8250],mmio32,<addr>[,options]
1175 uart[8250],mmio32be,<addr>[,options]
1176 uart[8250],0x<addr>[,options]
1177 Start an early, polled-mode console on the 8250/16550
1178 UART at the specified I/O port or MMIO address.
1179 MMIO inter-register address stride is either 8-bit
1180 (mmio) or 32-bit (mmio32 or mmio32be).
1181 If none of [io|mmio|mmio32|mmio32be], <addr> is assumed
1182 to be equivalent to 'mmio'. 'options' are specified
1183 in the same format described for "console=ttyS<n>"; if
1184 unspecified, the h/w is not initialized.
1188 Start an early, polled-mode console on a pl011 serial
1189 port at the specified address. The pl011 serial port
1190 must already be setup and configured. Options are not
1191 yet supported. If 'mmio32' is specified, then only
1192 the driver will use only 32-bit accessors to read/write
1193 the device registers.
1196 Start an early console on a litex serial port at the
1197 specified address. The serial port must already be
1198 setup and configured. Options are not yet supported.
1201 Start an early, polled-mode console on a meson serial
1202 port at the specified address. The serial port must
1203 already be setup and configured. Options are not yet
1207 Start an early, polled-mode console on an msm serial
1208 port at the specified address. The serial port
1209 must already be setup and configured. Options are not
1212 msm_serial_dm,<addr>
1213 Start an early, polled-mode console on an msm serial
1214 dm port at the specified address. The serial port
1215 must already be setup and configured. Options are not
1219 Start an early, polled-mode console on a serial port
1220 of an Actions Semi SoC, such as S500 or S900, at the
1221 specified address. The serial port must already be
1222 setup and configured. Options are not yet supported.
1225 Start an early, polled-mode console on a serial port
1226 of an RDA Micro SoC, such as RDA8810PL, at the
1227 specified address. The serial port must already be
1228 setup and configured. Options are not yet supported.
1231 Use RISC-V SBI (Supervisor Binary Interface) for early
1234 smh Use ARM semihosting calls for early console.
1242 Use early console provided by serial driver available
1243 on Samsung SoCs, requires selecting proper type and
1244 a correct base address of the selected UART port. The
1245 serial port must already be setup and configured.
1246 Options are not yet supported.
1249 Start an early, polled-mode console on a lantiq serial
1250 (lqasc) port at the specified address. The serial port
1251 must already be setup and configured. Options are not
1256 Use early console provided by Freescale LP UART driver
1257 found on Freescale Vybrid and QorIQ LS1021A processors.
1258 A valid base address must be provided, and the serial
1259 port must already be setup and configured.
1263 Start an early, polled-mode, output-only console on the
1264 Freescale i.MX UART at the specified address. The UART
1265 must already be setup and configured.
1268 Start an early, polled-mode console on the
1269 Armada 3700 serial port at the specified
1270 address. The serial port must already be setup
1271 and configured. Options are not yet supported.
1274 Start an early, polled-mode console on a Qualcomm
1275 Generic Interface (GENI) based serial port at the
1276 specified address. The serial port must already be
1277 setup and configured. Options are not yet supported.
1280 Start an early, unaccelerated console on the EFI
1281 memory mapped framebuffer (if available). On cache
1282 coherent non-x86 systems that use system memory for
1283 the framebuffer, pass the 'ram' option so that it is
1284 mapped with the correct attributes.
1287 Use early console provided by Freescale LINFlexD UART
1288 serial driver for NXP S32V234 SoCs. A valid base
1289 address must be provided, and the serial port must
1290 already be setup and configured.
1292 earlyprintk= [X86,SH,ARM,M68k,S390]
1296 earlyprintk=serial[,ttySn[,baudrate]]
1297 earlyprintk=serial[,0x...[,baudrate]]
1298 earlyprintk=ttySn[,baudrate]
1299 earlyprintk=dbgp[debugController#]
1300 earlyprintk=pciserial[,force],bus:device.function[,baudrate]
1301 earlyprintk=xdbc[xhciController#]
1303 earlyprintk is useful when the kernel crashes before
1304 the normal console is initialized. It is not enabled by
1305 default because it has some cosmetic problems.
1307 Append ",keep" to not disable it when the real console
1310 Only one of vga, serial, or usb debug port can
1313 Currently only ttyS0 and ttyS1 may be specified by
1314 name. Other I/O ports may be explicitly specified
1315 on some architectures (x86 and arm at least) by
1316 replacing ttySn with an I/O port address, like this:
1317 earlyprintk=serial,0x1008,115200
1318 You can find the port for a given device in
1319 /proc/tty/driver/serial:
1320 2: uart:ST16650V2 port:00001008 irq:18 ...
1322 Interaction with the standard serial driver is not
1325 The VGA output is eventually overwritten by
1328 The xen option can only be used in Xen domains.
1330 The sclp output can only be used on s390.
1332 The optional "force" to "pciserial" enables use of a
1333 PCI device even when its classcode is not of the
1336 edac_report= [HW,EDAC] Control how to report EDAC event
1337 Format: {"on" | "off" | "force"}
1338 on: enable EDAC to report H/W event. May be overridden
1339 by other higher priority error reporting module.
1340 off: disable H/W event reporting through EDAC.
1341 force: enforce the use of EDAC to report H/W event.
1345 Format: {"off" | "on" | "skip[mbr]"}
1348 Format: { "debug", "disable_early_pci_dma",
1349 "nochunk", "noruntime", "nosoftreserve",
1350 "novamap", "no_disable_early_pci_dma" }
1351 debug: enable misc debug output.
1352 disable_early_pci_dma: disable the busmaster bit on all
1353 PCI bridges while in the EFI boot stub.
1354 nochunk: disable reading files in "chunks" in the EFI
1355 boot stub, as chunking can cause problems with some
1356 firmware implementations.
1357 noruntime : disable EFI runtime services support
1358 nosoftreserve: The EFI_MEMORY_SP (Specific Purpose)
1359 attribute may cause the kernel to reserve the
1360 memory range for a memory mapping driver to
1361 claim. Specify efi=nosoftreserve to disable this
1362 reservation and treat the memory by its base type
1363 (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM").
1364 novamap: do not call SetVirtualAddressMap().
1365 no_disable_early_pci_dma: Leave the busmaster bit set
1366 on all PCI bridges while in the EFI boot stub
1368 efi_no_storage_paranoia [EFI; X86]
1369 Using this parameter you can use more than 50% of
1370 your efi variable storage. Use this parameter only if
1371 you are really sure that your UEFI does sane gc and
1372 fulfills the spec otherwise your board may brick.
1374 efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86]
1375 Add arbitrary attribute to specific memory range by
1376 updating original EFI memory map.
1377 Region of memory which aa attribute is added to is
1380 If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000
1381 is specified, EFI_MEMORY_MORE_RELIABLE(0x10000)
1382 attribute is added to range 0x100000000-0x180000000 and
1383 0x10a0000000-0x1120000000.
1385 If efi_fake_mem=8G@9G:0x40000 is specified, the
1386 EFI_MEMORY_SP(0x40000) attribute is added to
1387 range 0x240000000-0x43fffffff.
1389 Using this parameter you can do debugging of EFI memmap
1390 related features. For example, you can do debugging of
1391 Address Range Mirroring feature even if your box
1392 doesn't support it, or mark specific memory as
1395 efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT
1396 that is to be dynamically loaded by Linux. If there are
1397 multiple variables with the same name but with different
1398 vendor GUIDs, all of them will be loaded. See
1399 Documentation/admin-guide/acpi/ssdt-overlays.rst for details.
1402 eisa_irq_edge= [PARISC,HW]
1403 See header of drivers/parisc/eisa.c.
1405 ekgdboc= [X86,KGDB] Allow early kernel console debugging
1408 This is designed to be used in conjunction with
1409 the boot argument: earlyprintk=vga
1411 This parameter works in place of the kgdboc parameter
1412 but can only be used if the backing tty is available
1413 very early in the boot process. For early debugging
1414 via a serial port see kgdboc_earlycon instead.
1417 See comment before function elanfreq_setup() in
1418 arch/x86/kernel/cpu/cpufreq/elanfreq.c.
1420 elfcorehdr=[size[KMG]@]offset[KMG] [IA64,PPC,SH,X86,S390]
1421 Specifies physical address of start of kernel core
1422 image elf header and optionally the size. Generally
1423 kexec loader will pass this option to capture kernel.
1424 See Documentation/admin-guide/kdump/kdump.rst for details.
1426 enable_mtrr_cleanup [X86]
1427 The kernel tries to adjust MTRR layout from continuous
1428 to discrete, to make X server driver able to add WB
1429 entry later. This parameter enables that.
1431 enable_timer_pin_1 [X86]
1432 Enable PIN 1 of APIC timer
1433 Can be useful to work around chipset bugs
1434 (in particular on some ATI chipsets).
1435 The kernel tries to set a reasonable default.
1437 enforcing [SELINUX] Set initial enforcing status.
1439 See security/selinux/Kconfig help text.
1440 0 -- permissive (log only, no denials).
1441 1 -- enforcing (deny and log).
1443 Value can be changed at runtime via
1444 /sys/fs/selinux/enforce.
1447 Disable Error Record Serialization Table (ERST)
1450 ether= [HW,NET] Ethernet cards parameters
1451 This option is obsoleted by the "netdev=" option, which
1452 has equivalent usage. See its documentation for details.
1456 Permit 'security.evm' to be updated regardless of
1457 current integrity status.
1462 fail_make_request=[KNL]
1463 General fault injection mechanism.
1464 Format: <interval>,<probability>,<space>,<times>
1465 See also Documentation/fault-injection/.
1468 Format: { initns | none }
1469 See Documentation/admin-guide/sysctl/net.rst for
1470 fb_tunnels_only_for_init_ns
1473 See Documentation/admin-guide/blockdev/floppy.rst.
1475 force_pal_cache_flush
1476 [IA-64] Avoid check_sal_cache_flush which may hang on
1477 buggy SAL_CACHE_FLUSH implementations. Using this
1478 parameter will force ia64_sal_cache_flush to call
1479 ia64_pal_cache_flush instead of SAL_CACHE_FLUSH.
1482 Forcefully enable Physical Address Extension (PAE).
1483 Many Pentium M systems disable PAE but may have a
1484 functionally usable PAE implementation.
1485 Warning: use of this parameter will taint the kernel
1486 and may cause unknown problems.
1489 [FTRACE] will set and start the specified tracer
1490 as early as possible in order to facilitate early
1493 ftrace_boot_snapshot
1494 [FTRACE] On boot up, a snapshot will be taken of the
1495 ftrace ring buffer that can be read at:
1496 /sys/kernel/tracing/snapshot.
1497 This is useful if you need tracing information from kernel
1498 boot up that is likely to be overridden by user space
1499 start up functionality.
1501 ftrace_dump_on_oops[=orig_cpu]
1502 [FTRACE] will dump the trace buffers on oops.
1503 If no parameter is passed, ftrace will dump
1504 buffers of all CPUs, but if you pass orig_cpu, it will
1505 dump only the buffer of the CPU that triggered the
1508 ftrace_filter=[function-list]
1509 [FTRACE] Limit the functions traced by the function
1510 tracer at boot up. function-list is a comma-separated
1511 list of functions. This list can be changed at run
1512 time by the set_ftrace_filter file in the debugfs
1515 ftrace_notrace=[function-list]
1516 [FTRACE] Do not trace the functions specified in
1517 function-list. This list can be changed at run time
1518 by the set_ftrace_notrace file in the debugfs
1521 ftrace_graph_filter=[function-list]
1522 [FTRACE] Limit the top level callers functions traced
1523 by the function graph tracer at boot up.
1524 function-list is a comma-separated list of functions
1525 that can be changed at run time by the
1526 set_graph_function file in the debugfs tracing directory.
1528 ftrace_graph_notrace=[function-list]
1529 [FTRACE] Do not trace from the functions specified in
1530 function-list. This list is a comma-separated list of
1531 functions that can be changed at run time by the
1532 set_graph_notrace file in the debugfs tracing directory.
1534 ftrace_graph_max_depth=<uint>
1535 [FTRACE] Used with the function graph tracer. This is
1536 the max depth it will trace into a function. This value
1537 can be changed at run time by the max_graph_depth file
1538 in the tracefs tracing directory. default: 0 (no limit)
1540 fw_devlink= [KNL] Create device links between consumer and supplier
1541 devices by scanning the firmware to infer the
1542 consumer/supplier relationships. This feature is
1543 especially useful when drivers are loaded as modules as
1544 it ensures proper ordering of tasks like device probing
1545 (suppliers first, then consumers), supplier boot state
1546 clean up (only after all consumers have probed),
1547 suspend/resume & runtime PM (consumers first, then
1549 Format: { off | permissive | on | rpm }
1550 off -- Don't create device links from firmware info.
1551 permissive -- Create device links from firmware info
1552 but use it only for ordering boot state clean
1553 up (sync_state() calls).
1554 on -- Create device links from firmware info and use it
1555 to enforce probe and suspend/resume ordering.
1556 rpm -- Like "on", but also use to order runtime PM.
1558 fw_devlink.strict=<bool>
1559 [KNL] Treat all inferred dependencies as mandatory
1560 dependencies. This only applies for fw_devlink=on|rpm.
1564 [HW,JOY] Multisystem joystick and NES/SNES/PSX pad
1565 support via parallel port (up to 5 devices per port)
1566 Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5>
1567 See also Documentation/input/devices/joystick-parport.rst
1571 gart_fix_e820= [X86-64] disable the fix e820 for K8 GART
1575 gcov_persist= [GCOV] When non-zero (default), profiling data for
1576 kernel modules is saved and remains accessible via
1577 debugfs, even when the module is unloaded/reloaded.
1578 When zero, profiling data is discarded and associated
1579 debugfs files are removed at module unload time.
1581 goldfish [X86] Enable the goldfish android emulator platform.
1582 Don't use this when you are not running on the
1585 gpio-mockup.gpio_mockup_ranges
1586 [HW] Sets the ranges of gpiochip of for this device.
1587 Format: <start1>,<end1>,<start2>,<end2>...
1588 gpio-mockup.gpio_mockup_named_lines
1589 [HW] Let the driver know GPIO lines should be named.
1591 gpt [EFI] Forces disk with valid GPT signature but
1592 invalid Protective MBR to be treated as GPT. If the
1593 primary GPT is corrupted, it enables the backup/alternate
1594 GPT to be used instead.
1596 grcan.enable0= [HW] Configuration of physical interface 0. Determines
1597 the "Enable 0" bit of the configuration register.
1600 grcan.enable1= [HW] Configuration of physical interface 1. Determines
1601 the "Enable 0" bit of the configuration register.
1604 grcan.select= [HW] Select which physical interface to use.
1607 grcan.txsize= [HW] Sets the size of the tx buffer.
1608 Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0.
1610 grcan.rxsize= [HW] Sets the size of the rx buffer.
1611 Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0.
1615 [KNL] Under CONFIG_HARDENED_USERCOPY, whether
1616 hardening is enabled for this boot. Hardened
1617 usercopy checking is used to protect the kernel
1618 from reading or writing beyond known memory
1619 allocation boundaries as a proactive defense
1620 against bounds-checking flaws in the kernel's
1621 copy_to_user()/copy_from_user() interface.
1622 on Perform hardened usercopy checks (default).
1623 off Disable hardened usercopy checks.
1625 hardlockup_all_cpu_backtrace=
1626 [KNL] Should the hard-lockup detector generate
1627 backtraces on all cpus.
1630 hashdist= [KNL,NUMA] Large hashes allocated during boot
1631 are distributed across NUMA nodes. Defaults on
1632 for 64-bit NUMA, off otherwise.
1633 Format: 0 | 1 (for off | on)
1635 hcl= [IA-64] SGI's Hardware Graph compatibility layer
1637 hd= [EIDE] (E)IDE hard drive subsystem geometry
1638 Format: <cyl>,<head>,<sect>
1641 Disable Hardware Error Source Table (HEST) support;
1642 corresponding firmware-first mode error processing
1643 logic will be disabled.
1645 hibernate= [HIBERNATION]
1646 noresume Don't check if there's a hibernation image
1647 present during boot.
1648 nocompress Don't compress/decompress hibernation images.
1649 no Disable hibernation and resume.
1650 protect_image Turn on image protection during restoration
1651 (that will set all pages holding image data
1652 during restoration read-only).
1654 highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact
1655 size of <nn>. This works even on boxes that have no
1656 highmem otherwise. This also works to reduce highmem
1657 size on bigger boxes.
1659 highres= [KNL] Enable/disable high resolution timer mode.
1660 Valid parameters: "on", "off"
1665 hpet= [X86-32,HPET] option to control HPET usage
1666 Format: { enable (default) | disable | force |
1668 disable: disable HPET and use PIT instead
1669 force: allow force enabled of undocumented chips (ICH4,
1671 verbose: show contents of HPET registers during setup
1673 hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET
1674 registers. Default set by CONFIG_HPET_MMAP_DEFAULT.
1676 hugepages= [HW] Number of HugeTLB pages to allocate at boot.
1677 If this follows hugepagesz (below), it specifies
1678 the number of pages of hugepagesz to be allocated.
1679 If this is the first HugeTLB parameter on the command
1680 line, it specifies the number of pages to allocate for
1681 the default huge page size. If using node format, the
1682 number of pages to allocate per-node can be specified.
1683 See also Documentation/admin-guide/mm/hugetlbpage.rst.
1684 Format: <integer> or (node format)
1685 <node>:<integer>[,<node>:<integer>]
1688 [HW] The size of the HugeTLB pages. This is used in
1689 conjunction with hugepages (above) to allocate huge
1690 pages of a specific size at boot. The pair
1691 hugepagesz=X hugepages=Y can be specified once for
1692 each supported huge page size. Huge page sizes are
1693 architecture dependent. See also
1694 Documentation/admin-guide/mm/hugetlbpage.rst.
1697 hugetlb_cma= [HW,CMA] The size of a CMA area used for allocation
1698 of gigantic hugepages. Or using node format, the size
1699 of a CMA area per node can be specified.
1700 Format: nn[KMGTPE] or (node format)
1701 <node>:nn[KMGTPE][,<node>:nn[KMGTPE]]
1703 Reserve a CMA area of given size and allocate gigantic
1704 hugepages using the CMA allocator. If enabled, the
1705 boot-time allocation of gigantic hugepages is skipped.
1707 hugetlb_free_vmemmap=
1708 [KNL] Reguires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
1710 Allows heavy hugetlb users to free up some more
1711 memory (7 * PAGE_SIZE for each 2MB hugetlb page).
1712 Format: { [oO][Nn]/Y/y/1 | [oO][Ff]/N/n/0 (default) }
1714 [oO][Nn]/Y/y/1: enable the feature
1715 [oO][Ff]/N/n/0: disable the feature
1717 Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y,
1720 This is not compatible with memory_hotplug.memmap_on_memory.
1721 If both parameters are enabled, hugetlb_free_vmemmap takes
1722 precedence over memory_hotplug.memmap_on_memory.
1725 [KNL] Should the hung task detector generate panics.
1728 A value of 1 instructs the kernel to panic when a
1729 hung task is detected. The default value is controlled
1730 by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time
1731 option. The value selected by this boot parameter can
1732 be changed later by the kernel.hung_task_panic sysctl.
1734 hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC)
1735 terminal devices. Valid values: 0..8
1736 hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs.
1737 If specified, z/VM IUCV HVC accepts connections
1738 from listed z/VM user IDs only.
1740 hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations
1741 which allow the hypervisor to 'idle' the
1742 guest on lock contention.
1745 Do not unregister boot console at start. This is only
1746 useful for debugging when something happens in the window
1747 between unregistering the boot console and initializing
1750 i2c_bus= [HW] Override the default board specific I2C bus speed
1751 or register an additional I2C bus that is not
1752 registered from board initialization code.
1756 i8042.debug [HW] Toggle i8042 debug mode
1757 i8042.unmask_kbd_data
1758 [HW] Enable printing of interrupt data from the KBD port
1759 (disabled by default, and as a pre-condition
1760 requires that i8042.debug=1 be enabled)
1761 i8042.direct [HW] Put keyboard port into non-translated mode
1762 i8042.dumbkbd [HW] Pretend that controller can only read data from
1763 keyboard and cannot control its state
1764 (Don't attempt to blink the leds)
1765 i8042.noaux [HW] Don't check for auxiliary (== mouse) port
1766 i8042.nokbd [HW] Don't check/create keyboard port
1767 i8042.noloop [HW] Disable the AUX Loopback command while probing
1769 i8042.nomux [HW] Don't check presence of an active multiplexing
1771 i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX
1773 i8042.notimeout [HW] Ignore timeout condition signalled by controller
1774 i8042.reset [HW] Reset the controller during init, cleanup and
1775 suspend-to-ram transitions, only during s2r
1776 transitions, or never reset
1777 Format: { 1 | Y | y | 0 | N | n }
1778 1, Y, y: always reset controller
1779 0, N, n: don't ever reset controller
1780 Default: only on s2r transitions on x86; most other
1781 architectures force reset to be always executed
1782 i8042.unlock [HW] Unlock (ignore) the keylock
1783 i8042.kbdreset [HW] Reset device connected to KBD port
1785 [HW] Allow deferred probing upon i8042 probe errors
1789 i915.invert_brightness=
1790 [DRM] Invert the sense of the variable that is used to
1791 set the brightness of the panel backlight. Normally a
1792 brightness value of 0 indicates backlight switched off,
1793 and the maximum of the brightness value sets the backlight
1794 to maximum brightness. If this parameter is set to 0
1795 (default) and the machine requires it, or this parameter
1796 is set to 1, a brightness value of 0 sets the backlight
1797 to maximum brightness, and the maximum of the brightness
1798 value switches the backlight off.
1799 -1 -- never invert brightness
1800 0 -- machine default
1801 1 -- force brightness inversion
1804 Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]]
1808 Format: idle=poll, idle=halt, idle=nomwait
1809 Poll forces a polling idle loop that can slightly
1810 improve the performance of waking up a idle CPU, but
1811 will use a lot of power and make the system run hot.
1813 idle=halt: Halt is forced to be used for CPU idle.
1814 In such case C2/C3 won't be used again.
1815 idle=nomwait: Disable mwait for CPU C-states
1819 Allow force disabling of Shared Virtual Memory (SVA)
1820 support for the idxd driver. By default it is set to
1823 idxd.tc_override= [HW]
1825 Allow override of default traffic class configuration
1826 for the device. By default it is set to false (0).
1828 ieee754= [MIPS] Select IEEE Std 754 conformance mode
1829 Format: { strict | legacy | 2008 | relaxed }
1832 Choose which programs will be accepted for execution
1833 based on the IEEE 754 NaN encoding(s) supported by
1834 the FPU and the NaN encoding requested with the value
1835 of an ELF file header flag individually set by each
1836 binary. Hardware implementations are permitted to
1837 support either or both of the legacy and the 2008 NaN
1840 Available settings are as follows:
1841 strict accept binaries that request a NaN encoding
1842 supported by the FPU
1843 legacy only accept legacy-NaN binaries, if supported
1845 2008 only accept 2008-NaN binaries, if supported
1847 relaxed accept any binaries regardless of whether
1848 supported by the FPU
1850 The FPU emulator is always able to support both NaN
1851 encodings, so if no FPU hardware is present or it has
1852 been disabled with 'nofpu', then the settings of
1853 'legacy' and '2008' strap the emulator accordingly,
1854 'relaxed' straps the emulator for both legacy-NaN and
1855 2008-NaN, whereas 'strict' enables legacy-NaN only on
1856 legacy processors and both NaN encodings on MIPS32 or
1859 The setting for ABS.fmt/NEG.fmt instruction execution
1860 mode generally follows that for the NaN encoding,
1861 except where unsupported by hardware.
1863 ignore_loglevel [KNL]
1864 Ignore loglevel setting - this will print /all/
1865 kernel messages to the console. Useful for debugging.
1866 We also add it as printk module parameter, so users
1867 could change it dynamically, usually by
1868 /sys/module/printk/parameters/ignore_loglevel.
1871 Ignore RLIMIT_DATA setting for data mappings,
1872 print warning at first misuse. Can be changed via
1873 /sys/module/kernel/parameters/ignore_rlimit_data.
1875 ihash_entries= [KNL]
1876 Set number of hash buckets for inode cache.
1878 ima_appraise= [IMA] appraise integrity measurements
1879 Format: { "off" | "enforce" | "fix" | "log" }
1882 ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead.
1883 The builtin appraise policy appraises all files
1886 ima_canonical_fmt [IMA]
1887 Use the canonical format for the binary runtime
1888 measurements, instead of host native format.
1891 Format: { md5 | sha1 | rmd160 | sha256 | sha384
1895 The list of supported hash algorithms is defined
1896 in crypto/hash_info.h.
1899 The builtin policies to load during IMA setup.
1900 Format: "tcb | appraise_tcb | secure_boot |
1901 fail_securely | critical_data"
1903 The "tcb" policy measures all programs exec'd, files
1904 mmap'd for exec, and all files opened with the read
1905 mode bit set by either the effective uid (euid=0) or
1908 The "appraise_tcb" policy appraises the integrity of
1909 all files owned by root.
1911 The "secure_boot" policy appraises the integrity
1912 of files (eg. kexec kernel image, kernel modules,
1913 firmware, policy, etc) based on file signatures.
1915 The "fail_securely" policy forces file signature
1916 verification failure also on privileged mounted
1917 filesystems with the SB_I_UNVERIFIABLE_SIGNATURE
1920 The "critical_data" policy measures kernel integrity
1923 ima_tcb [IMA] Deprecated. Use ima_policy= instead.
1924 Load a policy which meets the needs of the Trusted
1925 Computing Base. This means IMA will measure all
1926 programs exec'd, files mmap'd for exec, and all files
1927 opened for read by uid=0.
1930 Select one of defined IMA measurements template formats.
1931 Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" |
1936 [IMA] Define a custom template format.
1937 Format: { "field1|...|fieldN" }
1939 ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage
1940 Format: <min_file_size>
1941 Set the minimal file size for using asynchronous hash.
1942 If left unspecified, ahash usage is disabled.
1944 ahash performance varies for different data sizes on
1945 different crypto accelerators. This option can be used
1946 to achieve the best performance for a particular HW.
1948 ima.ahash_bufsize= [IMA] Asynchronous hash buffer size
1950 Set hashing buffer size. Default: 4k.
1952 ahash performance varies for different chunk sizes on
1953 different crypto accelerators. This option can be used
1954 to achieve best performance for particular HW.
1958 Run specified binary instead of /sbin/init as init
1961 initcall_debug [KNL] Trace initcalls as they are executed. Useful
1962 for working out where the kernel is dying during
1965 initcall_blacklist= [KNL] Do not execute a comma-separated list of
1966 initcall functions. Useful for debugging built-in
1967 modules and initcalls.
1969 initramfs_async= [KNL]
1972 This parameter controls whether the initramfs
1973 image is unpacked asynchronously, concurrently
1974 with devices being probed and
1975 initialized. This should normally just work,
1976 but as a debugging aid, one can get the
1977 historical behaviour of the initramfs
1978 unpacking being completed before device_ and
1981 initrd= [BOOT] Specify the location of the initial ramdisk
1983 initrdmem= [KNL] Specify a physical address and size from which to
1984 load the initrd. If an initrd is compiled in or
1985 specified in the bootparams, it takes priority over this
1987 Format: ss[KMG],nn[KMG]
1990 init_on_alloc= [MM] Fill newly allocated pages and heap objects with
1993 Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON.
1995 init_on_free= [MM] Fill freed pages and heap objects with zeroes.
1997 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON.
1999 init_pkru= [X86] Specify the default memory protection keys rights
2000 register contents for all processes. 0x55555554 by
2001 default (disallow access to all but pkey 0). Can
2002 override in debugfs after boot.
2004 inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver
2007 int_pln_enable [X86] Enable power limit notification interrupt
2009 integrity_audit=[IMA]
2010 Format: { "0" | "1" }
2011 0 -- basic integrity auditing messages. (Default)
2012 1 -- additional integrity auditing messages.
2014 intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option
2016 Enable intel iommu driver.
2018 Disable intel iommu driver.
2019 igfx_off [Default Off]
2020 By default, gfx is mapped as normal device. If a gfx
2021 device has a dedicated DMAR unit, the DMAR unit is
2022 bypassed by not enabling DMAR with this option. In
2023 this case, gfx device will use physical address for
2025 strict [Default Off]
2026 Deprecated, equivalent to iommu.strict=1.
2027 sp_off [Default Off]
2028 By default, super page will be supported if Intel IOMMU
2029 has the capability. With this option, super page will
2032 Enable the Intel IOMMU scalable mode if the hardware
2033 advertises that it has support for the scalable mode
2036 Disallow use of the Intel IOMMU scalable mode.
2037 tboot_noforce [Default Off]
2038 Do not force the Intel IOMMU enabled under tboot.
2039 By default, tboot will force Intel IOMMU on, which
2040 could harm performance of some high-throughput
2041 devices like 40GBit network cards, even if identity
2043 Note that using this option lowers the security
2044 provided by tboot because it makes the system
2045 vulnerable to DMA attacks.
2047 intel_idle.max_cstate= [KNL,HW,ACPI,X86]
2048 0 disables intel_idle and fall back on acpi_idle.
2049 1 to 9 specify maximum depth of C-state.
2053 Do not enable intel_pstate as the default
2054 scaling driver for the supported processors
2056 Use intel_pstate as a scaling driver, but configure it
2057 to work with generic cpufreq governors (instead of
2058 enabling its internal governor). This mode cannot be
2059 used along with the hardware-managed P-states (HWP)
2062 Enable intel_pstate on systems that prohibit it by default
2063 in favor of acpi-cpufreq. Forcing the intel_pstate driver
2064 instead of acpi-cpufreq may disable platform features, such
2065 as thermal controls and power capping, that rely on ACPI
2066 P-States information being indicated to OSPM and therefore
2067 should be used with caution. This option does not work with
2068 processors that aren't supported by the intel_pstate driver
2069 or on platforms that use pcc-cpufreq instead of acpi-cpufreq.
2071 Do not enable hardware P state control (HWP)
2074 Only load intel_pstate on systems which support
2075 hardware P state control (HWP) if available.
2077 Enforce ACPI _PPC performance limits. If the Fixed ACPI
2078 Description Table, specifies preferred power management
2079 profile as "Enterprise Server" or "Performance Server",
2080 then this feature is turned on by default.
2082 Allow per-logical-CPU P-State performance control limits using
2083 cpufreq sysfs interface
2085 intremap= [X86-64, Intel-IOMMU]
2086 on enable Interrupt Remapping (default)
2087 off disable Interrupt Remapping
2088 nosid disable Source ID checking
2090 BIOS x2APIC opt-out request will be ignored
2091 nopost disable Interrupt Posting
2093 iomem= Disable strict checking of access to MMIO memory
2094 strict regions from userspace.
2109 nobypass [PPC/POWERNV]
2110 Disable IOMMU bypass, using IOMMU for PCI devices.
2112 iommu.forcedac= [ARM64, X86] Control IOVA allocation for PCI devices.
2113 Format: { "0" | "1" }
2114 0 - Try to allocate a 32-bit DMA address first, before
2115 falling back to the full range if needed.
2116 1 - Allocate directly from the full usable range,
2117 forcing Dual Address Cycle for PCI cards supporting
2118 greater than 32-bit addressing.
2120 iommu.strict= [ARM64, X86] Configure TLB invalidation behaviour
2121 Format: { "0" | "1" }
2123 Request that DMA unmap operations use deferred
2124 invalidation of hardware TLBs, for increased
2125 throughput at the cost of reduced device isolation.
2126 Will fall back to strict mode if not supported by
2127 the relevant IOMMU driver.
2129 DMA unmap operations invalidate IOMMU hardware TLBs
2131 unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}.
2132 Note: on x86, strict mode specified via one of the
2133 legacy driver-specific options takes precedence.
2136 [ARM64, X86] Configure DMA to bypass the IOMMU by default.
2137 Format: { "0" | "1" }
2138 0 - Use IOMMU translation for DMA.
2139 1 - Bypass the IOMMU for DMA.
2140 unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH.
2142 io7= [HW] IO7 for Marvel-based Alpha systems
2143 See comment before marvel_specify_io7 in
2144 arch/alpha/kernel/core_marvel.c.
2146 io_delay= [X86] I/O delay method
2148 Standard port 0x80 based delay
2150 Alternate port 0xed based delay (needed on some systems)
2152 Simple two microseconds delay
2157 See Documentation/admin-guide/nfs/nfsroot.rst.
2159 ipcmni_extend [KNL] Extend the maximum number of unique System V
2160 IPC identifiers from 32,768 to 16,777,216.
2162 irqaffinity= [SMP] Set the default irq affinity mask
2163 The argument is a cpu list, as described above.
2165 irqchip.gicv2_force_probe=
2168 Force the kernel to look for the second 4kB page
2169 of a GICv2 controller even if the memory range
2170 exposed by the device tree is too small.
2172 irqchip.gicv3_nolpi=
2174 Force the kernel to ignore the availability of
2175 LPIs (and by consequence ITSs). Intended for system
2176 that use the kernel as a bootloader, and thus want
2177 to let secondary kernels in charge of setting up
2180 irqchip.gicv3_pseudo_nmi= [ARM64]
2181 Enables support for pseudo-NMIs in the kernel. This
2182 requires the kernel to be built with
2183 CONFIG_ARM64_PSEUDO_NMI.
2186 When an interrupt is not handled search all handlers
2187 for it. Intended to get systems with badly broken
2191 When an interrupt is not handled search all handlers
2192 for it. Also check all handlers each timer
2193 interrupt. Intended to get systems with badly broken
2197 Format: <RDP>,<reset>,<pci_scan>,<verbosity>
2199 isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
2200 [Deprecated - use cpusets instead]
2201 Format: [flag-list,]<cpu-list>
2203 Specify one or more CPUs to isolate from disturbances
2204 specified in the flag list (default: domain):
2207 Disable the tick when a single task runs.
2209 A residual 1Hz tick is offloaded to workqueues, which you
2210 need to affine to housekeeping through the global
2211 workqueue's affinity configured via the
2212 /sys/devices/virtual/workqueue/cpumask sysfs file, or
2213 by using the 'domain' flag described below.
2215 NOTE: by default the global workqueue runs on all CPUs,
2216 so to protect individual CPUs the 'cpumask' file has to
2217 be configured manually after bootup.
2220 Isolate from the general SMP balancing and scheduling
2221 algorithms. Note that performing domain isolation this way
2222 is irreversible: it's not possible to bring back a CPU to
2223 the domains once isolated through isolcpus. It's strongly
2224 advised to use cpusets instead to disable scheduler load
2225 balancing through the "cpuset.sched_load_balance" file.
2226 It offers a much more flexible interface where CPUs can
2227 move in and out of an isolated set anytime.
2229 You can move a process onto or off an "isolated" CPU via
2230 the CPU affinity syscalls or cpuset.
2231 <cpu number> begins at 0 and the maximum value is
2232 "number of CPUs in system - 1".
2236 Isolate from being targeted by managed interrupts
2237 which have an interrupt mask containing isolated
2238 CPUs. The affinity of managed interrupts is
2239 handled by the kernel and cannot be changed via
2240 the /proc/irq/* interfaces.
2242 This isolation is best effort and only effective
2243 if the automatically assigned interrupt mask of a
2244 device queue contains isolated and housekeeping
2245 CPUs. If housekeeping CPUs are online then such
2246 interrupts are directed to the housekeeping CPU
2247 so that IO submitted on the housekeeping CPU
2248 cannot disturb the isolated CPU.
2250 If a queue's affinity mask contains only isolated
2251 CPUs then this parameter has no effect on the
2252 interrupt routing decision, though interrupts are
2253 only delivered when tasks running on those
2254 isolated CPUs submit IO. IO submitted on
2255 housekeeping CPUs has no influence on those
2258 The format of <cpu-list> is described above.
2262 ivrs_ioapic [HW,X86-64]
2263 Provide an override to the IOAPIC-ID<->DEVICE-ID
2264 mapping provided in the IVRS ACPI table. For
2265 example, to map IOAPIC-ID decimal 10 to
2266 PCI device 00:14.0 write the parameter as:
2267 ivrs_ioapic[10]=00:14.0
2269 ivrs_hpet [HW,X86-64]
2270 Provide an override to the HPET-ID<->DEVICE-ID
2271 mapping provided in the IVRS ACPI table. For
2272 example, to map HPET-ID decimal 0 to
2273 PCI device 00:14.0 write the parameter as:
2274 ivrs_hpet[0]=00:14.0
2276 ivrs_acpihid [HW,X86-64]
2277 Provide an override to the ACPI-HID:UID<->DEVICE-ID
2278 mapping provided in the IVRS ACPI table. For
2279 example, to map UART-HID:UID AMD0020:0 to
2280 PCI device 00:14.5 write the parameter as:
2281 ivrs_acpihid[00:14.5]=AMD0020:0
2283 js= [HW,JOY] Analog joystick
2284 See Documentation/input/joydev/joystick.rst.
2287 When CONFIG_RANDOMIZE_BASE is set, this disables
2288 kernel and module base offset ASLR (Address Space
2289 Layout Randomization).
2292 [KNL] Enforce KASAN (Kernel Address Sanitizer) to print
2293 report on every invalid memory access. Without this
2294 parameter KASAN will print report only for the first
2299 kernelcore= [KNL,X86,IA-64,PPC]
2300 Format: nn[KMGTPE] | nn% | "mirror"
2301 This parameter specifies the amount of memory usable by
2302 the kernel for non-movable allocations. The requested
2303 amount is spread evenly throughout all nodes in the
2304 system as ZONE_NORMAL. The remaining memory is used for
2305 movable memory in its own zone, ZONE_MOVABLE. In the
2306 event, a node is too small to have both ZONE_NORMAL and
2307 ZONE_MOVABLE, kernelcore memory will take priority and
2308 other nodes will have a larger ZONE_MOVABLE.
2310 ZONE_MOVABLE is used for the allocation of pages that
2311 may be reclaimed or moved by the page migration
2312 subsystem. Note that allocations like PTEs-from-HighMem
2313 still use the HighMem zone if it exists, and the Normal
2314 zone if it does not.
2316 It is possible to specify the exact amount of memory in
2317 the form of "nn[KMGTPE]", a percentage of total system
2318 memory in the form of "nn%", or "mirror". If "mirror"
2319 option is specified, mirrored (reliable) memory is used
2320 for non-movable allocations and remaining memory is used
2321 for Movable pages. "nn[KMGTPE]", "nn%", and "mirror"
2322 are exclusive, so you cannot specify multiple forms.
2324 kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port.
2325 Format: <Controller#>[,poll interval]
2326 The controller # is the number of the ehci usb debug
2327 port as it is probed via PCI. The poll interval is
2328 optional and is the number seconds in between
2329 each poll cycle to the debug port in case you need
2330 the functionality for interrupting the kernel with
2331 gdb or control-c on the dbgp connection. When
2332 not using this parameter you use sysrq-g to break into
2333 the kernel debugger.
2335 kgdboc= [KGDB,HW] kgdb over consoles.
2336 Requires a tty driver that supports console polling,
2337 or a supported polling keyboard driver (non-usb).
2338 Serial only format: <serial_device>[,baud]
2339 keyboard only format: kbd
2340 keyboard and serial format: kbd,<serial_device>[,baud]
2341 Optional Kernel mode setting:
2342 kms, kbd format: kms,kbd
2343 kms, kbd and serial format: kms,kbd,<ser_dev>[,baud]
2345 kgdboc_earlycon= [KGDB,HW]
2346 If the boot console provides the ability to read
2347 characters and can work in polling mode, you can use
2348 this parameter to tell kgdb to use it as a backend
2349 until the normal console is registered. Intended to
2350 be used together with the kgdboc parameter which
2351 specifies the normal console to transition to.
2353 The name of the early console should be specified
2354 as the value of this parameter. Note that the name of
2355 the early console might be different than the tty
2356 name passed to kgdboc. It's OK to leave the value
2357 blank and the first boot console that implements
2358 read() will be picked.
2360 kgdbwait [KGDB] Stop kernel execution and enter the
2361 kernel debugger at the earliest opportunity.
2363 kmac= [MIPS] Korina ethernet MAC address.
2364 Configure the RouterBoard 532 series on-chip
2365 Ethernet adapter MAC address.
2367 kmemleak= [KNL] Boot-time kmemleak enable/disable
2368 Valid arguments: on, off
2370 Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y,
2373 kprobe_event=[probe-list]
2374 [FTRACE] Add kprobe events and enable at boot time.
2375 The probe-list is a semicolon delimited list of probe
2376 definitions. Each definition is same as kprobe_events
2377 interface, but the parameters are comma delimited.
2378 For example, to add a kprobe event on vfs_read with
2379 arg1 and arg2, add to the command line;
2381 kprobe_event=p,vfs_read,$arg1,$arg2
2383 See also Documentation/trace/kprobetrace.rst "Kernel
2384 Boot Parameter" section.
2386 kpti= [ARM64] Control page table isolation of user
2387 and kernel address spaces.
2388 Default: enabled on cores which need mitigation.
2392 kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs.
2393 Default is 0 (don't ignore, but inject #GP)
2395 kvm.eager_page_split=
2396 [KVM,X86] Controls whether or not KVM will try to
2397 proactively split all huge pages during dirty logging.
2398 Eager page splitting reduces interruptions to vCPU
2399 execution by eliminating the write-protection faults
2400 and MMU lock contention that would otherwise be
2401 required to split huge pages lazily.
2403 VM workloads that rarely perform writes or that write
2404 only to a small region of VM memory may benefit from
2405 disabling eager page splitting to allow huge pages to
2406 still be used for reads.
2408 The behavior of eager page splitting depends on whether
2409 KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If
2410 disabled, all huge pages in a memslot will be eagerly
2411 split when dirty logging is enabled on that memslot. If
2412 enabled, eager page splitting will be performed during
2413 the KVM_CLEAR_DIRTY ioctl, and only for the pages being
2416 Eager page splitting currently only supports splitting
2417 huge pages mapped by the TDP MMU.
2421 kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
2422 Default is false (don't support).
2425 [KVM] Controls the software workaround for the
2426 X86_BUG_ITLB_MULTIHIT bug.
2427 force : Always deploy workaround.
2428 off : Never deploy workaround.
2429 auto : Deploy workaround based on the presence of
2430 X86_BUG_ITLB_MULTIHIT.
2434 If the software workaround is enabled for the host,
2435 guests do need not to enable it for nested guests.
2437 kvm.nx_huge_pages_recovery_ratio=
2438 [KVM] Controls how many 4KiB pages are periodically zapped
2439 back to huge pages. 0 disables the recovery, otherwise if
2440 the value is N KVM will zap 1/Nth of the 4KiB pages every
2441 period (see below). The default is 60.
2443 kvm.nx_huge_pages_recovery_period_ms=
2444 [KVM] Controls the time period at which KVM zaps 4KiB pages
2445 back to huge pages. If the value is a non-zero N, KVM will
2446 zap a portion (see ratio above) of the pages every N msecs.
2447 If the value is 0 (the default), KVM will pick a period based
2448 on the ratio, such that a page is zapped after 1 hour on average.
2450 kvm-amd.nested= [KVM,AMD] Allow nested virtualization in KVM/SVM.
2451 Default is 1 (enabled)
2453 kvm-amd.npt= [KVM,AMD] Disable nested paging (virtualized MMU)
2455 Default is 1 (enabled) if in 64-bit or 32-bit PAE mode.
2458 [KVM,ARM] Select one of KVM/arm64's modes of operation.
2460 none: Forcefully disable KVM.
2462 nvhe: Standard nVHE-based mode, without support for
2465 protected: nVHE-based mode with support for guests whose
2466 state is kept private from the host.
2467 Not valid if the kernel is running in EL2.
2469 Defaults to VHE/nVHE based on hardware support. Setting
2470 mode to "protected" will disable kexec and hibernation
2473 kvm-arm.vgic_v3_group0_trap=
2474 [KVM,ARM] Trap guest accesses to GICv3 group-0
2477 kvm-arm.vgic_v3_group1_trap=
2478 [KVM,ARM] Trap guest accesses to GICv3 group-1
2481 kvm-arm.vgic_v3_common_trap=
2482 [KVM,ARM] Trap guest accesses to GICv3 common
2485 kvm-arm.vgic_v4_enable=
2486 [KVM,ARM] Allow use of GICv4 for direct injection of
2489 kvm_cma_resv_ratio=n [PPC]
2490 Reserves given percentage from system memory area for
2491 contiguous memory allocation for KVM hash pagetable
2493 By default it reserves 5% of total system memory.
2497 kvm-intel.ept= [KVM,Intel] Disable extended page tables
2498 (virtualized MMU) support on capable Intel chips.
2499 Default is 1 (enabled)
2501 kvm-intel.emulate_invalid_guest_state=
2502 [KVM,Intel] Disable emulation of invalid guest state.
2503 Ignored if kvm-intel.enable_unrestricted_guest=1, as
2504 guest state is never invalid for unrestricted guests.
2505 This param doesn't apply to nested guests (L2), as KVM
2506 never emulates invalid L2 guest state.
2507 Default is 1 (enabled)
2509 kvm-intel.flexpriority=
2510 [KVM,Intel] Disable FlexPriority feature (TPR shadow).
2511 Default is 1 (enabled)
2514 [KVM,Intel] Enable VMX nesting (nVMX).
2515 Default is 0 (disabled)
2517 kvm-intel.unrestricted_guest=
2518 [KVM,Intel] Disable unrestricted guest feature
2519 (virtualized real and unpaged mode) on capable
2520 Intel chips. Default is 1 (enabled)
2522 kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault
2525 Valid arguments: never, cond, always
2527 always: L1D cache flush on every VMENTER.
2528 cond: Flush L1D on VMENTER only when the code between
2529 VMEXIT and VMENTER can leak host memory.
2530 never: Disables the mitigation
2532 Default is cond (do L1 cache flush in specific instances)
2534 kvm-intel.vpid= [KVM,Intel] Disable Virtual Processor Identification
2535 feature (tagged TLBs) on capable Intel chips.
2536 Default is 1 (enabled)
2538 l1d_flush= [X86,INTEL]
2539 Control mitigation for L1D based snooping vulnerability.
2541 Certain CPUs are vulnerable to an exploit against CPU
2542 internal buffers which can forward information to a
2543 disclosure gadget under certain conditions.
2545 In vulnerable processors, the speculatively
2546 forwarded data can be used in a cache side channel
2547 attack, to access data to which the attacker does
2548 not have direct access.
2550 This parameter controls the mitigation. The
2553 on - enable the interface for the mitigation
2555 l1tf= [X86] Control mitigation of the L1TF vulnerability on
2558 The kernel PTE inversion protection is unconditionally
2559 enabled and cannot be disabled.
2562 Provides all available mitigations for the
2563 L1TF vulnerability. Disables SMT and
2564 enables all mitigations in the
2565 hypervisors, i.e. unconditional L1D flush.
2567 SMT control and L1D flush control via the
2568 sysfs interface is still possible after
2569 boot. Hypervisors will issue a warning
2570 when the first VM is started in a
2571 potentially insecure configuration,
2572 i.e. SMT enabled or L1D flush disabled.
2575 Same as 'full', but disables SMT and L1D
2576 flush runtime control. Implies the
2577 'nosmt=force' command line option.
2578 (i.e. sysfs control of SMT is disabled.)
2581 Leaves SMT enabled and enables the default
2582 hypervisor mitigation, i.e. conditional
2585 SMT control and L1D flush control via the
2586 sysfs interface is still possible after
2587 boot. Hypervisors will issue a warning
2588 when the first VM is started in a
2589 potentially insecure configuration,
2590 i.e. SMT enabled or L1D flush disabled.
2594 Disables SMT and enables the default
2595 hypervisor mitigation.
2597 SMT control and L1D flush control via the
2598 sysfs interface is still possible after
2599 boot. Hypervisors will issue a warning
2600 when the first VM is started in a
2601 potentially insecure configuration,
2602 i.e. SMT enabled or L1D flush disabled.
2605 Same as 'flush', but hypervisors will not
2606 warn when a VM is started in a potentially
2607 insecure configuration.
2610 Disables hypervisor mitigations and doesn't
2612 It also drops the swap size and available
2613 RAM limit restriction on both hypervisor and
2618 For details see: Documentation/admin-guide/hw-vuln/l1tf.rst
2624 lapic [X86-32,APIC] Enable the local APIC even if BIOS
2627 lapic= [X86,APIC] Do not use TSC deadline
2628 value for LAPIC timer one-shot implementation. Default
2629 back to the programmable timer unit in the LAPIC.
2630 Format: notscdeadline
2632 lapic_timer_c2_ok [X86,APIC] trust the local apic timer
2635 libata.dma= [LIBATA] DMA control
2636 libata.dma=0 Disable all PATA and SATA DMA
2637 libata.dma=1 PATA and SATA Disk DMA only
2638 libata.dma=2 ATAPI (CDROM) DMA only
2639 libata.dma=4 Compact Flash DMA only
2640 Combinations also work, so libata.dma=3 enables DMA
2641 for disks and CDROMs, but not CFs.
2643 libata.ignore_hpa= [LIBATA] Ignore HPA limit
2644 libata.ignore_hpa=0 keep BIOS limits (default)
2645 libata.ignore_hpa=1 ignore limits, using full disk
2647 libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume
2651 libata.force= [LIBATA] Force configurations. The format is a comma-
2652 separated list of "[ID:]VAL" where ID is PORT[.DEVICE].
2653 PORT and DEVICE are decimal numbers matching port, link
2654 or device. Basically, it matches the ATA ID string
2655 printed on console by libata. If the whole ID part is
2656 omitted, the last PORT and DEVICE values are used. If
2657 ID hasn't been specified yet, the configuration applies
2658 to all ports, links and devices.
2660 If only DEVICE is omitted, the parameter applies to
2661 the port and all links and devices behind it. DEVICE
2662 number of 0 either selects the first device or the
2663 first fan-out link behind PMP device. It does not
2664 select the host link. DEVICE number of 15 selects the
2665 host link and device attached to it.
2667 The VAL specifies the configuration to force. As long
2668 as there is no ambiguity, shortcut notation is allowed.
2669 For example, both 1.5 and 1.5G would work for 1.5Gbps.
2670 The following configurations can be forced.
2672 * Cable type: 40c, 80c, short40c, unk, ign or sata.
2673 Any ID with matching PORT is used.
2675 * SATA link speed limit: 1.5Gbps or 3.0Gbps.
2677 * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7].
2678 udma[/][16,25,33,44,66,100,133] notation is also
2681 * nohrst, nosrst, norst: suppress hard, soft and both
2684 * rstonce: only attempt one reset during hot-unplug
2687 * [no]dbdelay: Enable or disable the extra 200ms delay
2688 before debouncing a link PHY and device presence
2691 * [no]ncq: Turn on or off NCQ.
2693 * [no]ncqtrim: Enable or disable queued DSM TRIM.
2695 * [no]ncqati: Enable or disable NCQ trim on ATI chipset.
2697 * [no]trim: Enable or disable (unqueued) TRIM.
2699 * trim_zero: Indicate that TRIM command zeroes data.
2701 * max_trim_128m: Set 128M maximum trim size limit.
2703 * [no]dma: Turn on or off DMA transfers.
2705 * atapi_dmadir: Enable ATAPI DMADIR bridge support.
2707 * atapi_mod16_dma: Enable the use of ATAPI DMA for
2708 commands that are not a multiple of 16 bytes.
2710 * [no]dmalog: Enable or disable the use of the
2711 READ LOG DMA EXT command to access logs.
2713 * [no]iddevlog: Enable or disable access to the
2714 identify device data log.
2716 * [no]logdir: Enable or disable access to the general
2717 purpose log directory.
2719 * max_sec_128: Set transfer size limit to 128 sectors.
2721 * max_sec_1024: Set or clear transfer size limit to
2724 * max_sec_lba48: Set or clear transfer size limit to
2727 * [no]lpm: Enable or disable link power management.
2729 * [no]setxfer: Indicate if transfer speed mode setting
2732 * dump_id: Dump IDENTIFY data.
2734 * disable: Disable this device.
2736 If there are multiple matching configurations changing
2737 the same attribute, the last one is used.
2739 load_ramdisk= [RAM] [Deprecated]
2741 lockd.nlm_grace_period=P [NFS] Assign grace period.
2744 lockd.nlm_tcpport=N [NFS] Assign TCP port.
2747 lockd.nlm_timeout=T [NFS] Assign timeout value.
2750 lockd.nlm_udpport=M [NFS] Assign UDP port.
2753 lockdown= [SECURITY]
2754 { integrity | confidentiality }
2755 Enable the kernel lockdown feature. If set to
2756 integrity, kernel features that allow userland to
2757 modify the running kernel are disabled. If set to
2758 confidentiality, kernel features that allow userland
2759 to extract confidential information from the kernel
2762 locktorture.nreaders_stress= [KNL]
2763 Set the number of locking read-acquisition kthreads.
2764 Defaults to being automatically set based on the
2765 number of online CPUs.
2767 locktorture.nwriters_stress= [KNL]
2768 Set the number of locking write-acquisition kthreads.
2770 locktorture.onoff_holdoff= [KNL]
2771 Set time (s) after boot for CPU-hotplug testing.
2773 locktorture.onoff_interval= [KNL]
2774 Set time (s) between CPU-hotplug operations, or
2775 zero to disable CPU-hotplug testing.
2777 locktorture.shuffle_interval= [KNL]
2778 Set task-shuffle interval (jiffies). Shuffling
2779 tasks allows some CPUs to go into dyntick-idle
2780 mode during the locktorture test.
2782 locktorture.shutdown_secs= [KNL]
2783 Set time (s) after boot system shutdown. This
2784 is useful for hands-off automated testing.
2786 locktorture.stat_interval= [KNL]
2787 Time (s) between statistics printk()s.
2789 locktorture.stutter= [KNL]
2790 Time (s) to stutter testing, for example,
2791 specifying five seconds causes the test to run for
2792 five seconds, wait for five seconds, and so on.
2793 This tests the locking primitive's ability to
2794 transition abruptly to and from idle.
2796 locktorture.torture_type= [KNL]
2797 Specify the locking implementation to test.
2799 locktorture.verbose= [KNL]
2800 Enable additional printk() statements.
2802 logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver
2805 loglevel= All Kernel Messages with a loglevel smaller than the
2806 console loglevel will be printed to the console. It can
2807 also be changed with klogd or other programs. The
2808 loglevels are defined as follows:
2810 0 (KERN_EMERG) system is unusable
2811 1 (KERN_ALERT) action must be taken immediately
2812 2 (KERN_CRIT) critical conditions
2813 3 (KERN_ERR) error conditions
2814 4 (KERN_WARNING) warning conditions
2815 5 (KERN_NOTICE) normal but significant condition
2816 6 (KERN_INFO) informational
2817 7 (KERN_DEBUG) debug-level messages
2819 log_buf_len=n[KMG] Sets the size of the printk ring buffer,
2820 in bytes. n must be a power of two and greater
2821 than the minimal size. The minimal size is defined
2822 by LOG_BUF_SHIFT kernel config parameter. There is
2823 also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter
2824 that allows to increase the default size depending on
2825 the number of CPUs. See init/Kconfig for more details.
2827 logo.nologo [FB] Disables display of the built-in Linux logo.
2828 This may be used to provide more screen space for
2829 kernel log messages and is useful when debugging
2830 kernel boot problems.
2832 lp=0 [LP] Specify parallel ports to use, e.g,
2833 lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses
2834 lp=reset first parallel port). 'lp=0' disables the
2835 lp=auto printer driver. 'lp=reset' (which can be
2836 specified in addition to the ports) causes
2837 attached printers to be reset. Using
2838 lp=port1,port2,... specifies the parallel ports
2839 to associate lp devices with, starting with
2840 lp0. A port specification may be 'none' to skip
2841 that lp device, or a parport name such as
2842 'parport0'. Specifying 'lp=auto' instead of a
2843 port specification list means that device IDs
2844 from each port should be examined, to see if
2845 an IEEE 1284-compliant printer is attached; if
2846 so, the driver will manage that printer.
2847 See also header of drivers/char/lp.c.
2850 Sets loops_per_jiffy to given constant, thus avoiding
2851 time-consuming boot-time autodetection (up to 250 ms per
2852 CPU). 0 enables autodetection (default). To determine
2853 the correct value for your kernel, boot with normal
2854 autodetection and see what value is printed. Note that
2855 on SMP systems the preset will be applied to all CPUs,
2856 which is likely to cause problems if your CPUs need
2857 significantly divergent settings. An incorrect value
2858 will cause delays in the kernel to be wrong, leading to
2859 unpredictable I/O errors and other breakage. Although
2860 unlikely, in the extreme case this might damage your
2864 Format: <io>,<irq>,<dma>
2866 lsm.debug [SECURITY] Enable LSM initialization debugging output.
2869 [SECURITY] Choose order of LSM initialization. This
2870 overrides CONFIG_LSM, and the "security=" parameter.
2872 machvec= [IA-64] Force the use of a particular machine-vector
2873 (machvec) in a generic kernel.
2874 Example: machvec=hpzx1
2876 machtype= [Loongson] Share the same kernel image file between
2877 different yeeloong laptops.
2878 Example: machtype=lemote-yeeloong-2f-7inch
2880 max_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory greater
2881 than or equal to this physical address is ignored.
2883 maxcpus= [SMP] Maximum number of processors that an SMP kernel
2884 will bring up during bootup. maxcpus=n : n >= 0 limits
2885 the kernel to bring up 'n' processors. Surely after
2886 bootup you can bring up the other plugged cpu by executing
2887 "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus
2888 only takes effect during system bootup.
2889 While n=0 is a special case, it is equivalent to "nosmp",
2890 which also disables the IO APIC.
2892 max_loop= [LOOP] The number of loop block devices that get
2893 (loop.max_loop) unconditionally pre-created at init time. The default
2894 number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead
2895 of statically allocating a predefined number, loop
2896 devices can be requested on-demand with the
2897 /dev/loop-control interface.
2899 mce [X86-32] Machine Check Exception
2901 mce=option [X86-64] See Documentation/x86/x86_64/boot-options.rst
2903 md= [HW] RAID subsystems devices and level
2904 See Documentation/admin-guide/md.rst.
2907 Format: <first>,<last>
2908 Specifies range of consoles to be captured by the MDA.
2911 Control mitigation for the Micro-architectural Data
2912 Sampling (MDS) vulnerability.
2914 Certain CPUs are vulnerable to an exploit against CPU
2915 internal buffers which can forward information to a
2916 disclosure gadget under certain conditions.
2918 In vulnerable processors, the speculatively
2919 forwarded data can be used in a cache side channel
2920 attack, to access data to which the attacker does
2921 not have direct access.
2923 This parameter controls the MDS mitigation. The
2926 full - Enable MDS mitigation on vulnerable CPUs
2927 full,nosmt - Enable MDS mitigation and disable
2928 SMT on vulnerable CPUs
2929 off - Unconditionally disable MDS mitigation
2931 On TAA-affected machines, mds=off can be prevented by
2932 an active TAA mitigation as both vulnerabilities are
2933 mitigated with the same mechanism so in order to disable
2934 this mitigation, you need to specify tsx_async_abort=off
2937 Not specifying this option is equivalent to
2940 For details see: Documentation/admin-guide/hw-vuln/mds.rst
2942 mem=nn[KMG] [HEXAGON] Set the memory size.
2943 Must be specified, otherwise memory size will be 0.
2945 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
2946 Amount of memory to be used in cases as follows:
2949 2 when the kernel is not able to see the whole system memory;
2950 3 memory that lies after 'mem=' boundary is excluded from
2951 the hypervisor, then assigned to KVM guests.
2952 4 to limit the memory available for kdump kernel.
2954 [ARC,MICROBLAZE] - the limit applies only to low memory,
2955 high memory is not affected.
2957 [ARM64] - only limits memory covered by the linear
2958 mapping. The NOMAP regions are not affected.
2960 [X86] Work as limiting max address. Use together
2961 with memmap= to avoid physical address space collisions.
2962 Without memmap= PCI devices could be placed at addresses
2963 belonging to unused RAM.
2965 Note that this only takes effects during boot time since
2966 in above case 3, memory may need be hot added after boot
2967 if system memory of hypervisor is not sufficient.
2970 [ARM,MIPS] - override the memory layout reported by
2972 Define a memory region of size nn[KMG] starting at
2974 Multiple different regions can be specified with
2975 multiple mem= parameters on the command line.
2977 mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel
2980 memblock=debug [KNL] Enable memblock debug messages.
2983 [KNL,SH] Allow user to override the default size for
2984 per-device physically contiguous DMA buffers.
2986 memhp_default_state=online/offline
2987 [KNL] Set the initial state for the memory hotplug
2988 onlining policy. If not specified, the default value is
2989 set according to the
2990 CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config
2992 See Documentation/admin-guide/mm/memory-hotplug.rst.
2994 memmap=exactmap [KNL,X86] Enable setting of an exact
2995 E820 memory map, as specified by the user.
2996 Such memmap=exactmap lines can be constructed based on
2997 BIOS output or other requirements. See the memmap=nn@ss
3000 memmap=nn[KMG]@ss[KMG]
3001 [KNL, X86, MIPS, XTENSA] Force usage of a specific region of memory.
3002 Region of memory to be used is from ss to ss+nn.
3003 If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG],
3004 which limits max address to nn[KMG].
3005 Multiple different regions can be specified,
3008 memmap=100M@2G,100M#3G,1G!1024G
3010 memmap=nn[KMG]#ss[KMG]
3011 [KNL,ACPI] Mark specific memory as ACPI data.
3012 Region of memory to be marked is from ss to ss+nn.
3014 memmap=nn[KMG]$ss[KMG]
3015 [KNL,ACPI] Mark specific memory as reserved.
3016 Region of memory to be reserved is from ss to ss+nn.
3017 Example: Exclude memory from 0x18690000-0x1869ffff
3018 memmap=64K$0x18690000
3020 memmap=0x10000$0x18690000
3021 Some bootloaders may need an escape character before '$',
3022 like Grub2, otherwise '$' and the following number
3025 memmap=nn[KMG]!ss[KMG]
3026 [KNL,X86] Mark specific memory as protected.
3027 Region of memory to be used, from ss to ss+nn.
3028 The memory region may be marked as e820 type 12 (0xc)
3029 and is NVDIMM or ADR memory.
3031 memmap=<size>%<offset>-<oldtype>+<newtype>
3032 [KNL,ACPI] Convert memory within the specified region
3033 from <oldtype> to <newtype>. If "-<oldtype>" is left
3034 out, the whole region will be marked as <newtype>,
3035 even if previously unavailable. If "+<newtype>" is left
3036 out, matching memory will be removed. Types are
3037 specified as e820 types, e.g., 1 = RAM, 2 = reserved,
3038 3 = ACPI, 12 = PRAM.
3040 memory_corruption_check=0/1 [X86]
3041 Some BIOSes seem to corrupt the first 64k of
3042 memory when doing things like suspend/resume.
3043 Setting this option will scan the memory
3044 looking for corruption. Enabling this will
3045 both detect corruption and prevent the kernel
3046 from using the memory being corrupted.
3047 However, its intended as a diagnostic tool; if
3048 repeatable BIOS-originated corruption always
3049 affects the same memory, you can use memmap=
3050 to prevent the kernel from using that memory.
3052 memory_corruption_check_size=size [X86]
3053 By default it checks for corruption in the low
3054 64k, making this memory unavailable for normal
3055 use. Use this parameter to scan for
3056 corruption in more or less memory.
3058 memory_corruption_check_period=seconds [X86]
3059 By default it checks for corruption every 60
3060 seconds. Use this parameter to check at some
3061 other rate. 0 disables periodic checking.
3063 memory_hotplug.memmap_on_memory
3064 [KNL,X86,ARM] Boolean flag to enable this feature.
3065 Format: {on | off (default)}
3066 When enabled, runtime hotplugged memory will
3067 allocate its internal metadata (struct pages)
3068 from the hotadded memory which will allow to
3069 hotadd a lot of memory without requiring
3070 additional memory to do so.
3071 This feature is disabled by default because it
3072 has some implication on large (e.g. GB)
3073 allocations in some configurations (e.g. small
3075 The state of the flag can be read in
3076 /sys/module/memory_hotplug/parameters/memmap_on_memory.
3077 Note that even when enabled, there are a few cases where
3078 the feature is not effective.
3080 This is not compatible with hugetlb_free_vmemmap. If
3081 both parameters are enabled, hugetlb_free_vmemmap takes
3082 precedence over memory_hotplug.memmap_on_memory.
3084 memtest= [KNL,X86,ARM,M68K,PPC,RISCV] Enable memtest
3086 default : 0 <disable>
3087 Specifies the number of memtest passes to be
3088 performed. Each pass selects another test
3089 pattern from a given set of patterns. Memtest
3090 fills the memory with this pattern, validates
3091 memory contents and reserves bad memory
3092 regions that are detected.
3094 mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control
3095 Valid arguments: on, off
3096 Default (depends on kernel configuration option):
3097 on (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=y)
3098 off (CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT=n)
3099 mem_encrypt=on: Activate SME
3100 mem_encrypt=off: Do not activate SME
3102 Refer to Documentation/virt/kvm/amd-memory-encryption.rst
3103 for details on when memory encryption can be activated.
3105 mem_sleep_default= [SUSPEND] Default system suspend mode:
3106 s2idle - Suspend-To-Idle
3107 shallow - Power-On Suspend or equivalent (if supported)
3108 deep - Suspend-To-RAM or equivalent (if supported)
3109 See Documentation/admin-guide/pm/sleep-states.rst.
3111 meye.*= [HW] Set MotionEye Camera parameters
3112 See Documentation/admin-guide/media/meye.rst.
3114 mfgpt_irq= [IA-32] Specify the IRQ to use for the
3115 Multi-Function General Purpose Timers on AMD Geode
3118 mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when
3119 the BIOS has incorrectly applied a workaround. TinyBIOS
3120 version 0.98 is known to be affected, 0.99 fixes the
3121 problem by letting the user disable the workaround.
3125 min_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory below this
3126 physical address is ignored.
3128 mini2440= [ARM,HW,KNL]
3129 Format:[0..2][b][c][t]
3131 MINI2440 configuration specification:
3132 0 - The attached screen is the 3.5" TFT
3133 1 - The attached screen is the 7" TFT
3134 2 - The VGA Shield is attached (1024x768)
3135 Leaving out the screen size parameter will not load
3136 the TFT driver, and the framebuffer will be left
3138 b - Enable backlight. The TFT backlight pin will be
3139 linked to the kernel VESA blanking code and a GPIO
3140 LED. This parameter is not necessary when using the
3142 c - Enable the s3c camera interface.
3143 t - Reserved for enabling touchscreen support. The
3144 touchscreen support is not enabled in the mainstream
3145 kernel as of 2.6.30, a preliminary port can be found
3146 in the "bleeding edge" mini2440 support kernel at
3147 https://repo.or.cz/w/linux-2.6/mini2440.git
3150 [X86,PPC,S390,ARM64] Control optional mitigations for
3151 CPU vulnerabilities. This is a set of curated,
3152 arch-independent options, each of which is an
3153 aggregation of existing arch-specific options.
3156 Disable all optional CPU mitigations. This
3157 improves system performance, but it may also
3158 expose users to several CPU vulnerabilities.
3159 Equivalent to: nopti [X86,PPC]
3161 nospectre_v1 [X86,PPC]
3163 nospectre_v2 [X86,PPC,S390,ARM64]
3164 spectre_v2_user=off [X86]
3165 spec_store_bypass_disable=off [X86,PPC]
3166 ssbd=force-off [ARM64]
3169 tsx_async_abort=off [X86]
3170 kvm.nx_huge_pages=off [X86]
3171 srbds=off [X86,INTEL]
3172 no_entry_flush [PPC]
3173 no_uaccess_flush [PPC]
3176 This does not have any effect on
3177 kvm.nx_huge_pages when
3178 kvm.nx_huge_pages=force.
3181 Mitigate all CPU vulnerabilities, but leave SMT
3182 enabled, even if it's vulnerable. This is for
3183 users who don't want to be surprised by SMT
3184 getting disabled across kernel upgrades, or who
3185 have other ways of avoiding SMT-based attacks.
3186 Equivalent to: (default behavior)
3189 Mitigate all CPU vulnerabilities, disabling SMT
3190 if needed. This is for users who always want to
3191 be fully mitigated, even if it means losing SMT.
3192 Equivalent to: l1tf=flush,nosmt [X86]
3193 mds=full,nosmt [X86]
3194 tsx_async_abort=full,nosmt [X86]
3197 [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this
3198 parameter allows control of the logging verbosity for
3199 the additional memory initialisation checks. A value
3200 of 0 disables mminit logging and a level of 4 will
3201 log everything. Information is printed at KERN_DEBUG
3202 so loglevel=8 may also need to be specified.
3205 [KNL] When CONFIG_MODULE_SIG is set, this means that
3206 modules without (valid) signatures will fail to load.
3207 Note that if CONFIG_MODULE_SIG_FORCE is set, that
3208 is always true, so this option does nothing.
3210 module_blacklist= [KNL] Do not load a comma-separated list of
3211 modules. Useful for debugging problem modules.
3214 [MOUSE] Maximum time between finger touching and
3215 leaving touchpad surface for touch to be considered
3216 a tap and be reported as a left button click (for
3217 touchpads working in absolute mode only).
3219 mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices
3220 reporting absolute coordinates, such as tablets
3221 mousedev.yres= [MOUSE] Vertical screen resolution, used for devices
3222 reporting absolute coordinates, such as tablets
3224 movablecore= [KNL,X86,IA-64,PPC]
3225 Format: nn[KMGTPE] | nn%
3226 This parameter is the complement to kernelcore=, it
3227 specifies the amount of memory used for migratable
3228 allocations. If both kernelcore and movablecore is
3229 specified, then kernelcore will be at *least* the
3230 specified value but may be more. If movablecore on its
3231 own is specified, the administrator must be careful
3232 that the amount of memory usable for all allocations
3235 movable_node [KNL] Boot-time switch to make hotplugable memory
3236 NUMA nodes to be movable. This means that the memory
3237 of such nodes will be usable only for movable
3238 allocations which rules out almost all kernel
3239 allocations. Use with caution!
3241 MTD_Partition= [MTD]
3242 Format: <name>,<region-number>,<size>,<offset>
3244 MTD_Region= [MTD] Format:
3245 <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>]
3248 See drivers/mtd/parsers/cmdlinepart.c
3251 ARM/S3C2412 JIVE boot control
3253 See arch/arm/mach-s3c/mach-jive.c
3255 mtouchusb.raw_coordinates=
3256 [HW] Make the MicroTouch USB driver use raw coordinates
3257 ('y', default) or cooked coordinates ('n')
3259 mtrr_chunk_size=nn[KMG] [X86]
3260 used for mtrr cleanup. It is largest continuous chunk
3261 that could hold holes aka. UC entries.
3263 mtrr_gran_size=nn[KMG] [X86]
3264 Used for mtrr cleanup. It is granularity of mtrr block.
3266 Large value could prevent small alignment from
3269 mtrr_spare_reg_nr=n [X86]
3271 Range: 0,7 : spare reg number
3273 Used for mtrr cleanup. It is spare mtrr entries number.
3274 Set to 2 or more if your graphical card needs more.
3276 multitce=off [PPC] This parameter disables the use of the pSeries
3277 firmware feature for updating multiple TCE entries
3280 n2= [NET] SDL Inc. RISCom/N2 synchronous serial card
3282 netdev= [NET] Network devices parameters
3283 Format: <irq>,<io>,<mem_start>,<mem_end>,<name>
3284 Note that mem_start is often overloaded to mean
3285 something different and driver-specific.
3286 This usage is only documented in each driver source
3289 netpoll.carrier_timeout=
3290 [NET] Specifies amount of time (in seconds) that
3291 netpoll should wait for a carrier. By default netpoll
3295 [NETFILTER] Enable connection tracking flow accounting
3296 0 to disable accounting
3297 1 to enable accounting
3300 nfsaddrs= [NFS] Deprecated. Use ip= instead.
3301 See Documentation/admin-guide/nfs/nfsroot.rst.
3303 nfsroot= [NFS] nfs root filesystem for disk-less boxes.
3304 See Documentation/admin-guide/nfs/nfsroot.rst.
3306 nfsrootdebug [NFS] enable nfsroot debugging messages.
3307 See Documentation/admin-guide/nfs/nfsroot.rst.
3309 nfs.callback_nr_threads=
3310 [NFSv4] set the total number of threads that the
3311 NFS client will assign to service NFSv4 callback
3314 nfs.callback_tcpport=
3315 [NFS] set the TCP port on which the NFSv4 callback
3316 channel should listen.
3319 [NFS] sets the pathname to the program which is used
3320 to update the NFS client cache entries.
3322 nfs.cache_getent_timeout=
3323 [NFS] sets the timeout after which an attempt to
3324 update a cache entry is deemed to have failed.
3326 nfs.idmap_cache_timeout=
3327 [NFS] set the maximum lifetime for idmapper cache
3331 [NFS] enable 64-bit inode numbers.
3332 If zero, the NFS client will fake up a 32-bit inode
3333 number for the readdir() and stat() syscalls instead
3334 of returning the full 64-bit number.
3335 The default is to return 64-bit inode numbers.
3337 nfs.max_session_cb_slots=
3338 [NFSv4.1] Sets the maximum number of session
3339 slots the client will assign to the callback
3340 channel. This determines the maximum number of
3341 callbacks the client will process in parallel for
3342 a particular server.
3344 nfs.max_session_slots=
3345 [NFSv4.1] Sets the maximum number of session slots
3346 the client will attempt to negotiate with the server.
3347 This limits the number of simultaneous RPC requests
3348 that the client can send to the NFSv4.1 server.
3349 Note that there is little point in setting this
3350 value higher than the max_tcp_slot_table_limit.
3352 nfs.nfs4_disable_idmapping=
3353 [NFSv4] When set to the default of '1', this option
3354 ensures that both the RPC level authentication
3355 scheme and the NFS level operations agree to use
3356 numeric uids/gids if the mount is using the
3357 'sec=sys' security flavour. In effect it is
3358 disabling idmapping, which can make migration from
3359 legacy NFSv2/v3 systems to NFSv4 easier.
3360 Servers that do not support this mode of operation
3361 will be autodetected by the client, and it will fall
3362 back to using the idmapper.
3363 To turn off this behaviour, set the value to '0'.
3365 [NFS4] Specify an additional fixed unique ident-
3366 ification string that NFSv4 clients can insert into
3367 their nfs_client_id4 string. This is typically a
3368 UUID that is generated at system install time.
3370 nfs.send_implementation_id =
3371 [NFSv4.1] Send client implementation identification
3372 information in exchange_id requests.
3373 If zero, no implementation identification information
3375 The default is to send the implementation identification
3378 nfs.recover_lost_locks =
3379 [NFSv4] Attempt to recover locks that were lost due
3380 to a lease timeout on the server. Please note that
3381 doing this risks data corruption, since there are
3382 no guarantees that the file will remain unchanged
3383 after the locks are lost.
3384 If you want to enable the kernel legacy behaviour of
3385 attempting to recover these locks, then set this
3387 The default parameter value of '0' causes the kernel
3388 not to attempt recovery of lost locks.
3390 nfs4.layoutstats_timer =
3391 [NFSv4.2] Change the rate at which the kernel sends
3392 layoutstats to the pNFS metadata server.
3394 Setting this to value to 0 causes the kernel to use
3395 whatever value is the default set by the layout
3396 driver. A non-zero value sets the minimum interval
3397 in seconds between layoutstats transmissions.
3399 nfsd.inter_copy_offload_enable =
3400 [NFSv4.2] When set to 1, the server will support
3401 server-to-server copies for which this server is
3402 the destination of the copy.
3404 nfsd.nfsd4_ssc_umount_timeout =
3405 [NFSv4.2] When used as the destination of a
3406 server-to-server copy, knfsd temporarily mounts
3407 the source server. It caches the mount in case
3408 it will be needed again, and discards it if not
3409 used for the number of milliseconds specified by
3412 nfsd.nfs4_disable_idmapping=
3413 [NFSv4] When set to the default of '1', the NFSv4
3414 server will return only numeric uids and gids to
3415 clients using auth_sys, and will accept numeric uids
3416 and gids from such clients. This is intended to ease
3417 migration from NFSv2/v3.
3420 nmi_backtrace.backtrace_idle [KNL]
3421 Dump stacks even of idle CPUs in response to an
3422 NMI stack-backtrace request.
3424 nmi_debug= [KNL,SH] Specify one or more actions to take
3425 when a NMI is triggered.
3426 Format: [state][,regs][,debounce][,die]
3428 nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels
3429 Format: [panic,][nopanic,][num]
3431 0 - turn hardlockup detector in nmi_watchdog off
3432 1 - turn hardlockup detector in nmi_watchdog on
3433 When panic is specified, panic when an NMI watchdog
3434 timeout occurs (or 'nopanic' to not panic on an NMI
3435 watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set)
3436 To disable both hard and soft lockup detectors,
3437 please see 'nowatchdog'.
3438 This is useful when you use a panic=... timeout and
3439 need the box quickly up again.
3441 These settings can be accessed at runtime via
3442 the nmi_watchdog and hardlockup_panic sysctls.
3444 no387 [BUGS=X86-32] Tells the kernel to use the 387 maths
3445 emulation library even if a 387 maths coprocessor
3448 no5lvl [X86-64] Disable 5-level paging mode. Forces
3449 kernel to use 4-level paging instead.
3451 nofsgsbase [X86] Disables FSGSBASE instructions.
3454 [HW] Never suspend the console
3455 Disable suspending of consoles during suspend and
3456 hibernate operations. Once disabled, debugging
3457 messages can reach various consoles while the rest
3458 of the system is being put to sleep (ie, while
3459 debugging driver suspend/resume hooks). This may
3460 not work reliably with all consoles, but is known
3461 to work with serial and VGA consoles.
3462 To facilitate more flexible debugging, we also add
3463 console_suspend, a printk module parameter to control
3464 it. Users could use console_suspend (usually
3465 /sys/module/printk/parameters/console_suspend) to
3466 turn on/off it dynamically.
3468 novmcoredd [KNL,KDUMP]
3469 Disable device dump. Device dump allows drivers to
3470 append dump data to vmcore so you can collect driver
3471 specified debug info. Drivers can append the data
3472 without any limit and this data is stored in memory,
3473 so this may cause significant memory stress. Disabling
3474 device dump can help save memory but the driver debug
3475 data will be no longer available. This parameter
3476 is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP
3479 noaliencache [MM, NUMA, SLAB] Disables the allocation of alien
3480 caches in the slab allocator. Saves per-node memory,
3481 but will impact performance.
3485 noaltinstr [S390] Disables alternative instructions patching
3486 (CPU alternatives feature).
3488 noapic [SMP,APIC] Tells the kernel to not make use of any
3489 IOAPICs that may be present in the system.
3491 noautogroup Disable scheduler automatic task group creation.
3493 nobats [PPC] Do not use BATs for mapping kernel lowmem
3494 on "Classic" PPC cores.
3498 nodsp [SH] Disable hardware DSP at boot time.
3500 noefi Disable EFI runtime services support.
3502 no_entry_flush [PPC] Don't flush the L1-D cache when entering the kernel.
3507 Disable SMAP (Supervisor Mode Access Prevention)
3508 even if it is supported by processor.
3511 Disable SMEP (Supervisor Mode Execution Prevention)
3512 even if it is supported by processor.
3515 This affects only 32-bit executables.
3516 noexec32=on: enable non-executable mappings (default)
3517 read doesn't imply executable mappings
3518 noexec32=off: disable non-executable mappings
3519 read implies executable mappings
3521 nofpu [MIPS,SH] Disable hardware FPU at boot time.
3523 nofxsr [BUGS=X86-32] Disables x86 floating point extended
3524 register save and restore. The kernel will only save
3525 legacy floating-point registers on task switch.
3527 nohugeiomap [KNL,X86,PPC,ARM64] Disable kernel huge I/O mappings.
3529 nohugevmalloc [PPC] Disable kernel huge vmalloc mappings.
3531 nosmt [KNL,S390] Disable symmetric multithreading (SMT).
3532 Equivalent to smt=1.
3534 [KNL,X86] Disable symmetric multithreading (SMT).
3535 nosmt=force: Force disable SMT, cannot be undone
3536 via the sysfs control file.
3538 nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1
3539 (bounds check bypass). With this option data leaks are
3540 possible in the system.
3542 nospectre_v2 [X86,PPC_FSL_BOOK3E,ARM64] Disable all mitigations for
3543 the Spectre variant 2 (indirect branch prediction)
3544 vulnerability. System may allow data leaks with this
3547 nospec_store_bypass_disable
3548 [HW] Disable all mitigations for the Speculative Store Bypass vulnerability
3551 [PPC] Don't flush the L1-D cache after accessing user data.
3553 noxsave [BUGS=X86] Disables x86 extended register state save
3554 and restore using xsave. The kernel will fallback to
3555 enabling legacy floating-point and sse state.
3557 noxsaveopt [X86] Disables xsaveopt used in saving x86 extended
3558 register states. The kernel will fall back to use
3559 xsave to save the states. By using this parameter,
3560 performance of saving the states is degraded because
3561 xsave doesn't support modified optimization while
3562 xsaveopt supports it on xsaveopt enabled systems.
3564 noxsaves [X86] Disables xsaves and xrstors used in saving and
3565 restoring x86 extended register state in compacted
3566 form of xsave area. The kernel will fall back to use
3567 xsaveopt and xrstor to save and restore the states
3568 in standard form of xsave area. By using this
3569 parameter, xsave area per process might occupy more
3570 memory on xsaves enabled systems.
3572 nohlt [ARM,ARM64,MICROBLAZE,SH] Forces the kernel to busy wait
3573 in do_idle() and not use the arch_cpu_idle()
3574 implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP
3575 to be effective. This is useful on platforms where the
3576 sleep(SH) or wfi(ARM,ARM64) instructions do not work
3577 correctly or when doing power measurements to evalute
3578 the impact of the sleep instructions. This is also
3579 useful when using JTAG debugger.
3581 no_file_caps Tells the kernel not to honor file capabilities. The
3582 only way then for a file to be executed with privilege
3583 is to be setuid root or executed by root.
3585 nohalt [IA-64] Tells the kernel not to use the power saving
3586 function PAL_HALT_LIGHT when idle. This increases
3587 power-consumption. On the positive side, it reduces
3588 interrupt wake-up latency, which may improve performance
3589 in certain environments such as networked servers or
3593 Force pointers printed to the console or buffers to be
3594 unhashed. By default, when a pointer is printed via %p
3595 format string, that pointer is "hashed", i.e. obscured
3596 by hashing the pointer value. This is a security feature
3597 that hides actual kernel addresses from unprivileged
3598 users, but it also makes debugging the kernel more
3599 difficult since unequal pointers can no longer be
3600 compared. However, if this command-line option is
3601 specified, then all normal pointers will have their true
3602 value printed. This option should only be specified when
3603 debugging the kernel. Please do not use on production
3606 nohibernate [HIBERNATION] Disable hibernation and resume.
3608 nohz= [KNL] Boottime enable/disable dynamic ticks
3609 Valid arguments: on, off
3612 nohz_full= [KNL,BOOT,SMP,ISOL]
3613 The argument is a cpu list, as described above.
3614 In kernels built with CONFIG_NO_HZ_FULL=y, set
3615 the specified list of CPUs whose tick will be stopped
3616 whenever possible. The boot CPU will be forced outside
3617 the range to maintain the timekeeping. Any CPUs
3618 in this list will have their RCU callbacks offloaded,
3619 just as if they had also been called out in the
3620 rcu_nocbs= boot parameter.
3622 noiotrap [SH] Disables trapped I/O port accesses.
3624 noirqdebug [X86-32] Disables the code which attempts to detect and
3625 disable unhandled interrupt sources.
3627 no_timer_check [X86,APIC] Disables the code which tests for
3628 broken timer IRQ sources.
3630 noisapnp [ISAPNP] Disables ISA PnP code.
3632 noinitrd [RAM] Tells the kernel not to load any configured
3635 nointremap [X86-64, Intel-IOMMU] Do not enable interrupt
3637 [Deprecated - use intremap=off]
3641 noinvpcid [X86] Disable the INVPCID cpu feature.
3643 nojitter [IA-64] Disables jitter checking for ITC timers.
3645 no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
3647 no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
3651 [X86,PV_OPS] Disable paravirtualized VMware scheduler
3652 clock and use the default one.
3654 no-steal-acc [X86,PV_OPS,ARM64] Disable paravirtualized steal time
3655 accounting. steal time is computed, but won't
3656 influence scheduler behaviour
3658 nolapic [X86-32,APIC] Do not enable or use the local APIC.
3660 nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
3662 noltlbs [PPC] Do not use large page/tlb entries for kernel
3663 lowmem mapping on PPC40x and PPC8xx
3665 nomca [IA-64] Disable machine check abort handling
3667 nomce [X86-32] Disable Machine Check Exception
3669 nomfgpt [X86-32] Disable Multi-Function General Purpose
3670 Timer usage (for AMD Geode machines).
3672 nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to
3673 shutdown the other cpus. Instead use the REBOOT_VECTOR
3676 nomodeset Disable kernel modesetting. DRM drivers will not perform
3677 display-mode changes or accelerated rendering. Only the
3678 system framebuffer will be available for use if this was
3679 set-up by the firmware or boot loader.
3681 Useful as fallback, or for testing and debugging.
3683 nomodule Disable module load
3685 nopat [X86] Disable PAT (page attribute table extension of
3686 pagetables) support.
3688 nopcid [X86-64] Disable the PCID cpu feature.
3690 norandmaps Don't use address space randomization. Equivalent to
3691 echo 0 > /proc/sys/kernel/randomize_va_space
3693 noreplace-smp [X86-32,SMP] Don't replace SMP instructions
3694 with UP alternatives
3696 nordrand [X86] Disable kernel use of the RDRAND and
3697 RDSEED instructions even if they are supported
3698 by the processor. RDRAND and RDSEED are still
3699 available to user space applications.
3701 noresume [SWSUSP] Disables resume and restores original swap
3704 no-scroll [VGA] Disables scrollback.
3705 This is required for the Braillex ib80-piezo Braille
3706 reader made by F.H. Papenmeier (Germany).
3710 nosgx [X86-64,SGX] Disables Intel SGX kernel support.
3712 nosmp [SMP] Tells an SMP kernel to act as a UP kernel,
3713 and disable the IO APIC. legacy for "maxcpus=0".
3715 nosoftlockup [KNL] Disable the soft-lockup detector.
3717 nosync [HW,M68K] Disables sync negotiation for all devices.
3719 nowatchdog [KNL] Disable both lockup detectors, i.e.
3720 soft-lockup and NMI watchdog (hard-lockup).
3724 nox2apic [X86-64,APIC] Do not enable x2APIC mode.
3726 nps_mtm_hs_ctr= [KNL,ARC]
3727 This parameter sets the maximum duration, in
3728 cycles, each HW thread of the CTOP can run
3729 without interruptions, before HW switches it.
3730 The actual maximum duration is 16 times this
3732 Format: integer between 1 and 255
3735 nptcg= [IA-64] Override max number of concurrent global TLB
3736 purges which is reported from either PAL_VM_SUMMARY or
3739 nr_cpus= [SMP] Maximum number of processors that an SMP kernel
3740 could support. nr_cpus=n : n >= 1 limits the kernel to
3741 support 'n' processors. It could be larger than the
3742 number of already plugged CPU during bootup, later in
3743 runtime you can physically add extra cpu until it reaches
3744 n. So during boot up some boot time memory for per-cpu
3745 variables need be pre-allocated for later physical cpu
3748 nr_uarts= [SERIAL] maximum number of UARTs to be registered.
3750 numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86] Disable NUMA, Only
3751 set up a single NUMA node spanning all memory.
3753 numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic
3755 Allowed values are enable and disable
3757 numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA.
3758 'node', 'default' can be specified
3759 This can be set from sysctl after boot.
3760 See Documentation/admin-guide/sysctl/vm.rst for details.
3762 ohci1394_dma=early [HW] enable debugging via the ohci1394 driver.
3763 See Documentation/core-api/debugging-via-ohci1394.rst for more
3766 olpc_ec_timeout= [OLPC] ms delay when issuing EC commands
3767 Rather than timing out after 20 ms if an EC
3768 command is not properly ACKed, override the length
3769 of the timeout. We have interrupts disabled while
3770 waiting for the ACK, so if this is set too high
3771 interrupts *may* be lost!
3773 omap_mux= [OMAP] Override bootloader pin multiplexing.
3774 Format: <mux_mode0.mode_name=value>...
3775 For example, to override I2C bus2:
3776 omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100
3778 onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration
3780 Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock]
3782 boundary - index of last SLC block on Flex-OneNAND.
3783 The remaining blocks are configured as MLC blocks.
3784 lock - Configure if Flex-OneNAND boundary should be locked.
3785 Once locked, the boundary cannot be changed.
3786 1 indicates lock status, 0 indicates unlock status.
3788 oops=panic Always panic on oopses. Default is to just kill the
3789 process, but there is a small probability of
3790 deadlocking the machine.
3791 This will also cause panics on machine check exceptions.
3792 Useful together with panic=30 to trigger a reboot.
3795 [KNL] Boolean flag to control whether the page allocator
3796 should randomize its free lists. The randomization may
3797 be automatically enabled if the kernel detects it is
3798 running on a platform with a direct-mapped memory-side
3799 cache, and this parameter can be used to
3800 override/disable that behavior. The state of the flag
3801 can be read from sysfs at:
3802 /sys/module/page_alloc/parameters/shuffle.
3804 page_owner= [KNL] Boot-time page_owner enabling option.
3805 Storage of the information about who allocated
3806 each page is disabled in default. With this switch,
3808 on: enable the feature
3810 page_poison= [KNL] Boot-time parameter changing the state of
3811 poisoning on the buddy allocator, available with
3812 CONFIG_PAGE_POISONING=y.
3813 off: turn off poisoning (default)
3814 on: turn on poisoning
3816 page_reporting.page_reporting_order=
3817 [KNL] Minimal page reporting order
3819 Adjust the minimal page reporting order. The page
3820 reporting is disabled when it exceeds (MAX_ORDER-1).
3822 panic= [KNL] Kernel behaviour on panic: delay <timeout>
3823 timeout > 0: seconds before rebooting
3824 timeout = 0: wait forever
3825 timeout < 0: reboot immediately
3828 panic_print= Bitmask for printing system info when panic happens.
3829 User can chose combination of the following bits:
3830 bit 0: print all tasks info
3831 bit 1: print system memory info
3832 bit 2: print timer info
3833 bit 3: print locks info if CONFIG_LOCKDEP is on
3834 bit 4: print ftrace buffer
3835 bit 5: print all printk messages in buffer
3836 bit 6: print all CPUs backtrace (if available in the arch)
3837 *Be aware* that this option may print a _lot_ of lines,
3838 so there are risks of losing older messages in the log.
3839 Use this option carefully, maybe worth to setup a
3840 bigger log buffer with "log_buf_len" along with this.
3842 panic_on_taint= Bitmask for conditionally calling panic() in add_taint()
3843 Format: <hex>[,nousertaint]
3844 Hexadecimal bitmask representing the set of TAINT flags
3845 that will cause the kernel to panic when add_taint() is
3846 called with any of the flags in this set.
3847 The optional switch "nousertaint" can be utilized to
3848 prevent userspace forced crashes by writing to sysctl
3849 /proc/sys/kernel/tainted any flagset matching with the
3850 bitmask set on panic_on_taint.
3851 See Documentation/admin-guide/tainted-kernels.rst for
3852 extra details on the taint flags that users can pick
3853 to compose the bitmask to assign to panic_on_taint.
3855 panic_on_warn panic() instead of WARN(). Useful to cause kdump
3858 parkbd.port= [HW] Parallel port number the keyboard adapter is
3859 connected to, default is 0.
3861 parkbd.mode= [HW] Parallel port keyboard adapter mode of operation,
3862 0 for XT, 1 for AT (default is AT).
3865 parport= [HW,PPT] Specify parallel ports. 0 disables.
3866 Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] }
3867 Use 'auto' to force the driver to use any
3868 IRQ/DMA settings detected (the default is to
3869 ignore detected IRQ/DMA settings because of
3870 possible conflicts). You can specify the base
3871 address, IRQ, and DMA settings; IRQ and DMA
3872 should be numbers, or 'auto' (for using detected
3873 settings on that particular port), or 'nofifo'
3874 (to avoid using a FIFO even if it is detected).
3875 Parallel ports are assigned in the order they
3876 are specified on the command line, starting
3879 parport_init_mode= [HW,PPT]
3880 Configure VIA parallel port to operate in
3881 a specific mode. This is necessary on Pegasos
3882 computer where firmware has no options for setting
3883 up parallel port mode and sets it to spp.
3884 Currently this function knows 686a and 8231 chips.
3885 Format: [spp|ps2|epp|ecp|ecpepp]
3887 pata_legacy.all= [HW,LIBATA]
3889 Set to non-zero to probe primary and secondary ISA
3890 port ranges on PCI systems where no PCI PATA device
3891 has been found at either range. Disabled by default.
3893 pata_legacy.autospeed= [HW,LIBATA]
3895 Set to non-zero if a chip is present that snoops speed
3896 changes. Disabled by default.
3898 pata_legacy.ht6560a= [HW,LIBATA]
3900 Set to 1, 2, or 3 for HT 6560A on the primary channel,
3901 the secondary channel, or both channels respectively.
3902 Disabled by default.
3904 pata_legacy.ht6560b= [HW,LIBATA]
3906 Set to 1, 2, or 3 for HT 6560B on the primary channel,
3907 the secondary channel, or both channels respectively.
3908 Disabled by default.
3910 pata_legacy.iordy_mask= [HW,LIBATA]
3912 IORDY enable mask. Set individual bits to allow IORDY
3913 for the respective channel. Bit 0 is for the first
3914 legacy channel handled by this driver, bit 1 is for
3915 the second channel, and so on. The sequence will often
3916 correspond to the primary legacy channel, the secondary
3917 legacy channel, and so on, but the handling of a PCI
3918 bus and the use of other driver options may interfere
3919 with the sequence. By default IORDY is allowed across
3922 pata_legacy.opti82c46x= [HW,LIBATA]
3924 Set to 1, 2, or 3 for Opti 82c611A on the primary
3925 channel, the secondary channel, or both channels
3926 respectively. Disabled by default.
3928 pata_legacy.opti82c611a= [HW,LIBATA]
3930 Set to 1, 2, or 3 for Opti 82c465MV on the primary
3931 channel, the secondary channel, or both channels
3932 respectively. Disabled by default.
3934 pata_legacy.pio_mask= [HW,LIBATA]
3936 PIO mode mask for autospeed devices. Set individual
3937 bits to allow the use of the respective PIO modes.
3938 Bit 0 is for mode 0, bit 1 is for mode 1, and so on.
3939 All modes allowed by default.
3941 pata_legacy.probe_all= [HW,LIBATA]
3943 Set to non-zero to probe tertiary and further ISA
3944 port ranges on PCI systems. Disabled by default.
3946 pata_legacy.probe_mask= [HW,LIBATA]
3948 Probe mask for legacy ISA PATA ports. Depending on
3949 platform configuration and the use of other driver
3950 options up to 6 legacy ports are supported: 0x1f0,
3951 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing
3952 of individual ports can be disabled by setting the
3953 corresponding bits in the mask to 1. Bit 0 is for
3954 the first port in the list above (0x1f0), and so on.
3955 By default all supported ports are probed.
3957 pata_legacy.qdi= [HW,LIBATA]
3959 Set to non-zero to probe QDI controllers. By default
3960 set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise.
3962 pata_legacy.winbond= [HW,LIBATA]
3964 Set to non-zero to probe Winbond controllers. Use
3965 the standard I/O port (0x130) if 1, otherwise the
3966 value given is the I/O port to use (typically 0x1b0).
3967 By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE,
3970 pata_platform.pio_mask= [HW,LIBATA]
3972 Supported PIO mode mask. Set individual bits to allow
3973 the use of the respective PIO modes. Bit 0 is for
3974 mode 0, bit 1 is for mode 1, and so on. Mode 0 only
3978 Halt all CPUs after the first oops has been printed for
3979 the specified number of seconds. This is to be used if
3980 your oopses keep scrolling off the screen.
3985 See header of drivers/block/paride/pcd.c.
3986 See also Documentation/admin-guide/blockdev/paride.rst.
3988 pci=option[,option...] [PCI] various PCI subsystem options.
3990 Some options herein operate on a specific device
3991 or a set of devices (<pci_dev>). These are
3992 specified in one of the following formats:
3994 [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]*
3995 pci:<vendor>:<device>[:<subvendor>:<subdevice>]
3997 Note: the first format specifies a PCI
3998 bus/device/function address which may change
3999 if new hardware is inserted, if motherboard
4000 firmware changes, or due to changes caused
4001 by other kernel parameters. If the
4002 domain is left unspecified, it is
4003 taken to be zero. Optionally, a path
4004 to a device through multiple device/function
4005 addresses can be specified after the base
4006 address (this is more robust against
4007 renumbering issues). The second format
4008 selects devices using IDs from the
4009 configuration space which may match multiple
4010 devices in the system.
4012 earlydump dump PCI config space before the kernel
4014 off [X86] don't probe for the PCI bus
4015 bios [X86-32] force use of PCI BIOS, don't access
4016 the hardware directly. Use this if your machine
4017 has a non-standard PCI host bridge.
4018 nobios [X86-32] disallow use of PCI BIOS, only direct
4019 hardware access methods are allowed. Use this
4020 if you experience crashes upon bootup and you
4021 suspect they are caused by the BIOS.
4022 conf1 [X86] Force use of PCI Configuration Access
4023 Mechanism 1 (config address in IO port 0xCF8,
4024 data in IO port 0xCFC, both 32-bit).
4025 conf2 [X86] Force use of PCI Configuration Access
4026 Mechanism 2 (IO port 0xCF8 is an 8-bit port for
4027 the function, IO port 0xCFA, also 8-bit, sets
4028 bus number. The config space is then accessed
4029 through ports 0xC000-0xCFFF).
4030 See http://wiki.osdev.org/PCI for more info
4031 on the configuration access mechanisms.
4032 noaer [PCIE] If the PCIEAER kernel config parameter is
4033 enabled, this kernel boot option can be used to
4034 disable the use of PCIE advanced error reporting.
4035 nodomains [PCI] Disable support for multiple PCI
4036 root domains (aka PCI segments, in ACPI-speak).
4037 nommconf [X86] Disable use of MMCONFIG for PCI
4039 check_enable_amd_mmconf [X86] check for and enable
4040 properly configured MMIO access to PCI
4041 config space on AMD family 10h CPU
4042 nomsi [MSI] If the PCI_MSI kernel config parameter is
4043 enabled, this kernel boot option can be used to
4044 disable the use of MSI interrupts system-wide.
4045 noioapicquirk [APIC] Disable all boot interrupt quirks.
4046 Safety option to keep boot IRQs enabled. This
4047 should never be necessary.
4048 ioapicreroute [APIC] Enable rerouting of boot IRQs to the
4049 primary IO-APIC for bridges that cannot disable
4050 boot IRQs. This fixes a source of spurious IRQs
4051 when the system masks IRQs.
4052 noioapicreroute [APIC] Disable workaround that uses the
4053 boot IRQ equivalent of an IRQ that connects to
4054 a chipset where boot IRQs cannot be disabled.
4055 The opposite of ioapicreroute.
4056 biosirq [X86-32] Use PCI BIOS calls to get the interrupt
4057 routing table. These calls are known to be buggy
4058 on several machines and they hang the machine
4059 when used, but on other computers it's the only
4060 way to get the interrupt routing table. Try
4061 this option if the kernel is unable to allocate
4062 IRQs or discover secondary PCI buses on your
4064 rom [X86] Assign address space to expansion ROMs.
4065 Use with caution as certain devices share
4066 address decoders between ROMs and other
4068 norom [X86] Do not assign address space to
4069 expansion ROMs that do not already have
4070 BIOS assigned address ranges.
4071 nobar [X86] Do not assign address space to the
4072 BARs that weren't assigned by the BIOS.
4073 irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be
4074 assigned automatically to PCI devices. You can
4075 make the kernel exclude IRQs of your ISA cards
4077 pirqaddr=0xAAAAA [X86] Specify the physical address
4078 of the PIRQ table (normally generated
4079 by the BIOS) if it is outside the
4080 F0000h-100000h range.
4081 lastbus=N [X86] Scan all buses thru bus #N. Can be
4082 useful if the kernel is unable to find your
4083 secondary buses and you want to tell it
4084 explicitly which ones they are.
4085 assign-busses [X86] Always assign all PCI bus
4086 numbers ourselves, overriding
4087 whatever the firmware may have done.
4088 usepirqmask [X86] Honor the possible IRQ mask stored
4089 in the BIOS $PIR table. This is needed on
4090 some systems with broken BIOSes, notably
4091 some HP Pavilion N5400 and Omnibook XE3
4092 notebooks. This will have no effect if ACPI
4093 IRQ routing is enabled.
4094 noacpi [X86] Do not use ACPI for IRQ routing
4095 or for PCI scanning.
4096 use_crs [X86] Use PCI host bridge window information
4097 from ACPI. On BIOSes from 2008 or later, this
4098 is enabled by default. If you need to use this,
4099 please report a bug.
4100 nocrs [X86] Ignore PCI host bridge windows from ACPI.
4101 If you need to use this, please report a bug.
4102 use_e820 [X86] Use E820 reservations to exclude parts of
4103 PCI host bridge windows. This is a workaround
4104 for BIOS defects in host bridge _CRS methods.
4105 If you need to use this, please report a bug to
4106 <linux-pci@vger.kernel.org>.
4107 no_e820 [X86] Ignore E820 reservations for PCI host
4108 bridge windows. This is the default on modern
4109 hardware. If you need to use this, please report
4110 a bug to <linux-pci@vger.kernel.org>.
4111 routeirq Do IRQ routing for all PCI devices.
4112 This is normally done in pci_enable_device(),
4113 so this option is a temporary workaround
4114 for broken drivers that don't call it.
4115 skip_isa_align [X86] do not align io start addr, so can
4116 handle more pci cards
4117 noearly [X86] Don't do any early type 1 scanning.
4118 This might help on some broken boards which
4119 machine check when some devices' config space
4120 is read. But various workarounds are disabled
4121 and some IOMMU drivers will not work.
4122 bfsort Sort PCI devices into breadth-first order.
4123 This sorting is done to get a device
4124 order compatible with older (<= 2.4) kernels.
4125 nobfsort Don't sort PCI devices into breadth-first order.
4126 pcie_bus_tune_off Disable PCIe MPS (Max Payload Size)
4127 tuning and use the BIOS-configured MPS defaults.
4128 pcie_bus_safe Set every device's MPS to the largest value
4129 supported by all devices below the root complex.
4130 pcie_bus_perf Set device MPS to the largest allowable MPS
4131 based on its parent bus. Also set MRRS (Max
4132 Read Request Size) to the largest supported
4133 value (no larger than the MPS that the device
4134 or bus can support) for best performance.
4135 pcie_bus_peer2peer Set every device's MPS to 128B, which
4136 every device is guaranteed to support. This
4137 configuration allows peer-to-peer DMA between
4138 any pair of devices, possibly at the cost of
4139 reduced performance. This also guarantees
4140 that hot-added devices will work.
4141 cbiosize=nn[KMG] The fixed amount of bus space which is
4142 reserved for the CardBus bridge's IO window.
4143 The default value is 256 bytes.
4144 cbmemsize=nn[KMG] The fixed amount of bus space which is
4145 reserved for the CardBus bridge's memory
4146 window. The default value is 64 megabytes.
4149 [<order of align>@]<pci_dev>[; ...]
4150 Specifies alignment and device to reassign
4151 aligned memory resources. How to
4152 specify the device is described above.
4153 If <order of align> is not specified,
4154 PAGE_SIZE is used as alignment.
4155 A PCI-PCI bridge can be specified if resource
4156 windows need to be expanded.
4157 To specify the alignment for several
4158 instances of a device, the PCI vendor,
4159 device, subvendor, and subdevice may be
4160 specified, e.g., 12@pci:8086:9c22:103c:198f
4161 for 4096-byte alignment.
4162 ecrc= Enable/disable PCIe ECRC (transaction layer
4163 end-to-end CRC checking).
4164 bios: Use BIOS/firmware settings. This is the
4168 hpiosize=nn[KMG] The fixed amount of bus space which is
4169 reserved for hotplug bridge's IO window.
4170 Default size is 256 bytes.
4171 hpmmiosize=nn[KMG] The fixed amount of bus space which is
4172 reserved for hotplug bridge's MMIO window.
4173 Default size is 2 megabytes.
4174 hpmmioprefsize=nn[KMG] The fixed amount of bus space which is
4175 reserved for hotplug bridge's MMIO_PREF window.
4176 Default size is 2 megabytes.
4177 hpmemsize=nn[KMG] The fixed amount of bus space which is
4178 reserved for hotplug bridge's MMIO and
4180 Default size is 2 megabytes.
4181 hpbussize=nn The minimum amount of additional bus numbers
4182 reserved for buses below a hotplug bridge.
4184 realloc= Enable/disable reallocating PCI bridge resources
4185 if allocations done by BIOS are too small to
4186 accommodate resources required by all child
4188 off: Turn realloc off
4190 realloc same as realloc=on
4191 noari do not use PCIe ARI.
4192 noats [PCIE, Intel-IOMMU, AMD-IOMMU]
4193 do not use PCIe ATS (and IOMMU device IOTLB).
4194 pcie_scan_all Scan all possible PCIe devices. Otherwise we
4195 only look for one device below a PCIe downstream
4197 big_root_window Try to add a big 64bit memory window to the PCIe
4198 root complex on AMD CPUs. Some GFX hardware
4199 can resize a BAR to allow access to all VRAM.
4200 Adding the window is slightly risky (it may
4201 conflict with unreported devices), so this
4203 disable_acs_redir=<pci_dev>[; ...]
4204 Specify one or more PCI devices (in the format
4205 specified above) separated by semicolons.
4206 Each device specified will have the PCI ACS
4207 redirect capabilities forced off which will
4208 allow P2P traffic between devices through
4209 bridges without forcing it upstream. Note:
4210 this removes isolation between devices and
4211 may put more devices in an IOMMU group.
4212 force_floating [S390] Force usage of floating interrupts.
4213 nomio [S390] Do not use MIO instructions.
4214 norid [S390] ignore the RID field and force use of
4215 one PCI domain per PCI function
4217 pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power
4220 force Enable ASPM even on devices that claim not to support it.
4221 WARNING: Forcing ASPM on may cause system lockups.
4223 pcie_ports= [PCIE] PCIe port services handling:
4224 native Use native PCIe services (PME, AER, DPC, PCIe hotplug)
4225 even if the platform doesn't give the OS permission to
4226 use them. This may cause conflicts if the platform
4227 also tries to use these services.
4228 dpc-native Use native PCIe service for DPC only. May
4229 cause conflicts if firmware uses AER or DPC.
4230 compat Disable native PCIe services (PME, AER, DPC, PCIe
4233 pcie_port_pm= [PCIE] PCIe port power management handling:
4234 off Disable power management of all PCIe ports
4235 force Forcibly enable power management of all PCIe ports
4237 pcie_pme= [PCIE,PM] Native PCIe PME signaling options:
4238 nomsi Do not use MSI for native PCIe PME signaling (this makes
4239 all PCIe root ports use INTx for all services).
4241 pcmv= [HW,PCMCIA] BadgePAD 4
4245 Keep all power-domains already enabled by bootloader on,
4246 even if no driver has claimed them. This is useful
4247 for debug and development, but should not be
4248 needed on a platform with proper driver support.
4251 See Documentation/admin-guide/blockdev/paride.rst.
4253 pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at
4256 See arch/parisc/kernel/pdc_chassis.c
4258 percpu_alloc= Select which percpu first chunk allocator to use.
4259 Currently supported values are "embed" and "page".
4260 Archs may support subset or none of the selections.
4261 See comments in mm/percpu.c for details on each
4262 allocator. This parameter is primarily for debugging
4263 and performance comparison.
4266 See Documentation/admin-guide/blockdev/paride.rst.
4269 See Documentation/admin-guide/blockdev/paride.rst.
4271 pirq= [SMP,APIC] Manual mp-table setup
4272 See Documentation/x86/i386/IO-APIC.rst.
4274 plip= [PPT,NET] Parallel port network link
4275 Format: { parport<nr> | timid | 0 }
4276 See also Documentation/admin-guide/parport.rst.
4278 pmtmr= [X86] Manual setup of pmtmr I/O Port.
4279 Override pmtimer IOPort with a hex value.
4282 pmu_override= [PPC] Override the PMU.
4283 This option takes over the PMU facility, so it is no
4284 longer usable by perf. Setting this option starts the
4285 PMU counters by setting MMCR0 to 0 (the FC bit is
4286 cleared). If a number is given, then MMCR1 is set to
4287 that number, otherwise (e.g., 'pmu_override=on'), MMCR1
4290 pm_debug_messages [SUSPEND,KNL]
4291 Enable suspend/resume debug messages during boot up.
4294 Enable PNP debug messages (depends on the
4295 CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time
4296 via /sys/module/pnp/parameters/debug. We always show
4297 current resource usage; turning this on also shows
4298 possible settings and some assignment information.
4304 { on | off | curr | res | no-curr | no-res }
4307 [ISAPNP] Exclude IRQs for the autoconfiguration
4310 [ISAPNP] Exclude DMAs for the autoconfiguration
4312 pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration
4313 Ranges are in pairs (I/O port base and size).
4316 [ISAPNP] Exclude memory regions for the
4318 Ranges are in pairs (memory base and size).
4320 ports= [IP_VS_FTP] IPVS ftp helper module
4322 Up to 8 (IP_VS_APP_MAX_PORTS) ports
4324 Format: <port>,<port>....
4326 powersave=off [PPC] This option disables power saving features.
4327 It specifically disables cpuidle and sets the
4328 platform machine description specific power_save
4329 function to NULL. On Idle the CPU just reduces
4332 ppc_strict_facility_enable
4333 [PPC] This option catches any kernel floating point,
4334 Altivec, VSX and SPE outside of regions specifically
4335 allowed (eg kernel_enable_fpu()/kernel_disable_fpu()).
4336 There is some performance impact when enabling this.
4340 Disable Hardware Transactional Memory
4343 Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC
4344 none - Limited to cond_resched() calls
4345 voluntary - Limited to cond_resched() and might_sleep() calls
4346 full - Any section that isn't explicitly preempt disabled
4347 can be preempted anytime.
4349 print-fatal-signals=
4350 [KNL] debug: print fatal signals
4352 If enabled, warn about various signal handling
4353 related application anomalies: too many signals,
4354 too many POSIX.1 timers, fatal signals causing a
4357 If you hit the warning due to signal overflow,
4358 you might want to try "ulimit -i unlimited".
4362 printk.always_kmsg_dump=
4363 Trigger kmsg_dump for cases other than kernel oops or
4365 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4368 printk.console_no_auto_verbose=
4369 Disable console loglevel raise on oops, panic
4370 or lockdep-detected issues (only if lock debug is on).
4371 With an exception to setups with low baudrate on
4372 serial console, keeping this 0 is a good choice
4373 in order to provide more debug information.
4375 default: 0 (auto_verbose is enabled)
4377 printk.devkmsg={on,off,ratelimit}
4378 Control writing to /dev/kmsg.
4379 on - unlimited logging to /dev/kmsg from userspace
4380 off - logging to /dev/kmsg disabled
4381 ratelimit - ratelimit the logging
4384 printk.time= Show timing data prefixed to each printk message line
4385 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4387 processor.max_cstate= [HW,ACPI]
4388 Limit processor to maximum C-state
4389 max_cstate=9 overrides any DMI blacklist limit.
4391 processor.nocst [HW,ACPI]
4392 Ignore the _CST method to determine C-states,
4393 instead using the legacy FADT method
4395 profile= [KNL] Enable kernel profiling via /proc/profile
4396 Format: [<profiletype>,]<number>
4397 Param: <profiletype>: "schedule", "sleep", or "kvm"
4398 [defaults to kernel profiling]
4399 Param: "schedule" - profile schedule points.
4400 Param: "sleep" - profile D-state sleeping (millisecs).
4401 Requires CONFIG_SCHEDSTATS
4402 Param: "kvm" - profile VM exits.
4403 Param: <number> - step/bucket size as a power of 2 for
4404 statistical time based profiling.
4406 prompt_ramdisk= [RAM] [Deprecated]
4408 prot_virt= [S390] enable hosting protected virtual machines
4409 isolated from the hypervisor (if hardware supports
4413 psi= [KNL] Enable or disable pressure stall information
4417 psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to
4418 probe for; one of (bare|imps|exps|lifebook|any).
4419 psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports
4421 psmouse.resetafter= [HW,MOUSE]
4422 Try to reset the device after so many bad packets
4425 [HW,MOUSE] Set desired mouse resolution, in dpi.
4426 psmouse.smartscroll=
4427 [HW,MOUSE] Controls Logitech smartscroll autorepeat.
4428 0 = disabled, 1 = enabled (default).
4430 pstore.backend= Specify the name of the pstore backend to use
4433 See Documentation/admin-guide/blockdev/paride.rst.
4435 pti= [X86-64] Control Page Table Isolation of user and
4436 kernel address spaces. Disabling this feature
4437 removes hardening, but improves performance of
4438 system calls and interrupts.
4440 on - unconditionally enable
4441 off - unconditionally disable
4442 auto - kernel detects whether your CPU model is
4443 vulnerable to issues that PTI mitigates
4445 Not specifying this option is equivalent to pti=auto.
4448 Equivalent to pti=off
4451 [KNL] Number of legacy pty's. Overwrites compiled-in
4454 quiet [KNL] Disable most log messages
4459 See Documentation/admin-guide/md.rst.
4461 ramdisk_size= [RAM] Sizes of RAM disks in kilobytes
4462 See Documentation/admin-guide/blockdev/ramdisk.rst.
4464 ramdisk_start= [RAM] RAM disk image start address
4466 random.trust_cpu={on,off}
4467 [KNL] Enable or disable trusting the use of the
4468 CPU's random number generator (if available) to
4469 fully seed the kernel's CRNG. Default is controlled
4470 by CONFIG_RANDOM_TRUST_CPU.
4472 random.trust_bootloader={on,off}
4473 [KNL] Enable or disable trusting the use of a
4474 seed passed by the bootloader (if available) to
4475 fully seed the kernel's CRNG. Default is controlled
4476 by CONFIG_RANDOM_TRUST_BOOTLOADER.
4478 randomize_kstack_offset=
4479 [KNL] Enable or disable kernel stack offset
4480 randomization, which provides roughly 5 bits of
4481 entropy, frustrating memory corruption attacks
4482 that depend on stack address determinism or
4483 cross-syscall address exposures. This is only
4484 available on architectures that have defined
4485 CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET.
4486 Format: <bool> (1/Y/y=enable, 0/N/n=disable)
4487 Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT.
4489 ras=option[,option,...] [KNL] RAS-specific options
4492 Disable the Correctable Errors Collector,
4493 see CONFIG_RAS_CEC help text.
4495 rcu_nocbs[=cpu-list]
4496 [KNL] The optional argument is a cpu list,
4499 In kernels built with CONFIG_RCU_NOCB_CPU=y,
4500 enable the no-callback CPU mode, which prevents
4501 such CPUs' callbacks from being invoked in
4502 softirq context. Invocation of such CPUs' RCU
4503 callbacks will instead be offloaded to "rcuox/N"
4504 kthreads created for that purpose, where "x" is
4505 "p" for RCU-preempt, "s" for RCU-sched, and "g"
4506 for the kthreads that mediate grace periods; and
4507 "N" is the CPU number. This reduces OS jitter on
4508 the offloaded CPUs, which can be useful for HPC
4509 and real-time workloads. It can also improve
4510 energy efficiency for asymmetric multiprocessors.
4512 If a cpulist is passed as an argument, the specified
4513 list of CPUs is set to no-callback mode from boot.
4515 Otherwise, if the '=' sign and the cpulist
4516 arguments are omitted, no CPU will be set to
4517 no-callback mode from boot but the mode may be
4518 toggled at runtime via cpusets.
4521 Rather than requiring that offloaded CPUs
4522 (specified by rcu_nocbs= above) explicitly
4523 awaken the corresponding "rcuoN" kthreads,
4524 make these kthreads poll for callbacks.
4525 This improves the real-time response for the
4526 offloaded CPUs by relieving them of the need to
4527 wake up the corresponding kthread, but degrades
4528 energy efficiency by requiring that the kthreads
4529 periodically wake up to do the polling.
4531 rcutree.blimit= [KNL]
4532 Set maximum number of finished RCU callbacks to
4533 process in one batch.
4535 rcutree.dump_tree= [KNL]
4536 Dump the structure of the rcu_node combining tree
4537 out at early boot. This is used for diagnostic
4538 purposes, to verify correct tree setup.
4540 rcutree.gp_cleanup_delay= [KNL]
4541 Set the number of jiffies to delay each step of
4542 RCU grace-period cleanup.
4544 rcutree.gp_init_delay= [KNL]
4545 Set the number of jiffies to delay each step of
4546 RCU grace-period initialization.
4548 rcutree.gp_preinit_delay= [KNL]
4549 Set the number of jiffies to delay each step of
4550 RCU grace-period pre-initialization, that is,
4551 the propagation of recent CPU-hotplug changes up
4552 the rcu_node combining tree.
4554 rcutree.use_softirq= [KNL]
4555 If set to zero, move all RCU_SOFTIRQ processing to
4556 per-CPU rcuc kthreads. Defaults to a non-zero
4557 value, meaning that RCU_SOFTIRQ is used by default.
4558 Specify rcutree.use_softirq=0 to use rcuc kthreads.
4560 But note that CONFIG_PREEMPT_RT=y kernels disable
4561 this kernel boot parameter, forcibly setting it
4564 rcutree.rcu_fanout_exact= [KNL]
4565 Disable autobalancing of the rcu_node combining
4566 tree. This is used by rcutorture, and might
4567 possibly be useful for architectures having high
4568 cache-to-cache transfer latencies.
4570 rcutree.rcu_fanout_leaf= [KNL]
4571 Change the number of CPUs assigned to each
4572 leaf rcu_node structure. Useful for very
4573 large systems, which will choose the value 64,
4574 and for NUMA systems with large remote-access
4575 latencies, which will choose a value aligned
4576 with the appropriate hardware boundaries.
4578 rcutree.rcu_min_cached_objs= [KNL]
4579 Minimum number of objects which are cached and
4580 maintained per one CPU. Object size is equal
4581 to PAGE_SIZE. The cache allows to reduce the
4582 pressure to page allocator, also it makes the
4583 whole algorithm to behave better in low memory
4586 rcutree.rcu_delay_page_cache_fill_msec= [KNL]
4587 Set the page-cache refill delay (in milliseconds)
4588 in response to low-memory conditions. The range
4589 of permitted values is in the range 0:100000.
4591 rcutree.jiffies_till_first_fqs= [KNL]
4592 Set delay from grace-period initialization to
4593 first attempt to force quiescent states.
4594 Units are jiffies, minimum value is zero,
4595 and maximum value is HZ.
4597 rcutree.jiffies_till_next_fqs= [KNL]
4598 Set delay between subsequent attempts to force
4599 quiescent states. Units are jiffies, minimum
4600 value is one, and maximum value is HZ.
4602 rcutree.jiffies_till_sched_qs= [KNL]
4603 Set required age in jiffies for a
4604 given grace period before RCU starts
4605 soliciting quiescent-state help from
4606 rcu_note_context_switch() and cond_resched().
4607 If not specified, the kernel will calculate
4608 a value based on the most recent settings
4609 of rcutree.jiffies_till_first_fqs
4610 and rcutree.jiffies_till_next_fqs.
4611 This calculated value may be viewed in
4612 rcutree.jiffies_to_sched_qs. Any attempt to set
4613 rcutree.jiffies_to_sched_qs will be cheerfully
4616 rcutree.kthread_prio= [KNL,BOOT]
4617 Set the SCHED_FIFO priority of the RCU per-CPU
4618 kthreads (rcuc/N). This value is also used for
4619 the priority of the RCU boost threads (rcub/N)
4620 and for the RCU grace-period kthreads (rcu_bh,
4621 rcu_preempt, and rcu_sched). If RCU_BOOST is
4622 set, valid values are 1-99 and the default is 1
4623 (the least-favored priority). Otherwise, when
4624 RCU_BOOST is not set, valid values are 0-99 and
4625 the default is zero (non-realtime operation).
4626 When RCU_NOCB_CPU is set, also adjust the
4627 priority of NOCB callback kthreads.
4629 rcutree.rcu_nocb_gp_stride= [KNL]
4630 Set the number of NOCB callback kthreads in
4631 each group, which defaults to the square root
4632 of the number of CPUs. Larger numbers reduce
4633 the wakeup overhead on the global grace-period
4634 kthread, but increases that same overhead on
4635 each group's NOCB grace-period kthread.
4637 rcutree.qhimark= [KNL]
4638 Set threshold of queued RCU callbacks beyond which
4639 batch limiting is disabled.
4641 rcutree.qlowmark= [KNL]
4642 Set threshold of queued RCU callbacks below which
4643 batch limiting is re-enabled.
4645 rcutree.qovld= [KNL]
4646 Set threshold of queued RCU callbacks beyond which
4647 RCU's force-quiescent-state scan will aggressively
4648 enlist help from cond_resched() and sched IPIs to
4649 help CPUs more quickly reach quiescent states.
4650 Set to less than zero to make this be set based
4651 on rcutree.qhimark at boot time and to zero to
4652 disable more aggressive help enlistment.
4654 rcutree.rcu_kick_kthreads= [KNL]
4655 Cause the grace-period kthread to get an extra
4656 wake_up() if it sleeps three times longer than
4657 it should at force-quiescent-state time.
4658 This wake_up() will be accompanied by a
4659 WARN_ONCE() splat and an ftrace_dump().
4661 rcutree.rcu_unlock_delay= [KNL]
4662 In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels,
4663 this specifies an rcu_read_unlock()-time delay
4664 in microseconds. This defaults to zero.
4665 Larger delays increase the probability of
4666 catching RCU pointer leaks, that is, buggy use
4667 of RCU-protected pointers after the relevant
4668 rcu_read_unlock() has completed.
4670 rcutree.sysrq_rcu= [KNL]
4671 Commandeer a sysrq key to dump out Tree RCU's
4672 rcu_node tree with an eye towards determining
4673 why a new grace period has not yet started.
4675 rcuscale.gp_async= [KNL]
4676 Measure performance of asynchronous
4677 grace-period primitives such as call_rcu().
4679 rcuscale.gp_async_max= [KNL]
4680 Specify the maximum number of outstanding
4681 callbacks per writer thread. When a writer
4682 thread exceeds this limit, it invokes the
4683 corresponding flavor of rcu_barrier() to allow
4684 previously posted callbacks to drain.
4686 rcuscale.gp_exp= [KNL]
4687 Measure performance of expedited synchronous
4688 grace-period primitives.
4690 rcuscale.holdoff= [KNL]
4691 Set test-start holdoff period. The purpose of
4692 this parameter is to delay the start of the
4693 test until boot completes in order to avoid
4696 rcuscale.kfree_rcu_test= [KNL]
4697 Set to measure performance of kfree_rcu() flooding.
4699 rcuscale.kfree_rcu_test_double= [KNL]
4700 Test the double-argument variant of kfree_rcu().
4701 If this parameter has the same value as
4702 rcuscale.kfree_rcu_test_single, both the single-
4703 and double-argument variants are tested.
4705 rcuscale.kfree_rcu_test_single= [KNL]
4706 Test the single-argument variant of kfree_rcu().
4707 If this parameter has the same value as
4708 rcuscale.kfree_rcu_test_double, both the single-
4709 and double-argument variants are tested.
4711 rcuscale.kfree_nthreads= [KNL]
4712 The number of threads running loops of kfree_rcu().
4714 rcuscale.kfree_alloc_num= [KNL]
4715 Number of allocations and frees done in an iteration.
4717 rcuscale.kfree_loops= [KNL]
4718 Number of loops doing rcuscale.kfree_alloc_num number
4719 of allocations and frees.
4721 rcuscale.nreaders= [KNL]
4722 Set number of RCU readers. The value -1 selects
4723 N, where N is the number of CPUs. A value
4724 "n" less than -1 selects N-n+1, where N is again
4725 the number of CPUs. For example, -2 selects N
4726 (the number of CPUs), -3 selects N+1, and so on.
4727 A value of "n" less than or equal to -N selects
4730 rcuscale.nwriters= [KNL]
4731 Set number of RCU writers. The values operate
4732 the same as for rcuscale.nreaders.
4733 N, where N is the number of CPUs
4735 rcuscale.perf_type= [KNL]
4736 Specify the RCU implementation to test.
4738 rcuscale.shutdown= [KNL]
4739 Shut the system down after performance tests
4740 complete. This is useful for hands-off automated
4743 rcuscale.verbose= [KNL]
4744 Enable additional printk() statements.
4746 rcuscale.writer_holdoff= [KNL]
4747 Write-side holdoff between grace periods,
4748 in microseconds. The default of zero says
4751 rcutorture.fqs_duration= [KNL]
4752 Set duration of force_quiescent_state bursts
4755 rcutorture.fqs_holdoff= [KNL]
4756 Set holdoff time within force_quiescent_state bursts
4759 rcutorture.fqs_stutter= [KNL]
4760 Set wait time between force_quiescent_state bursts
4763 rcutorture.fwd_progress= [KNL]
4764 Specifies the number of kthreads to be used
4765 for RCU grace-period forward-progress testing
4766 for the types of RCU supporting this notion.
4767 Defaults to 1 kthread, values less than zero or
4768 greater than the number of CPUs cause the number
4771 rcutorture.fwd_progress_div= [KNL]
4772 Specify the fraction of a CPU-stall-warning
4773 period to do tight-loop forward-progress testing.
4775 rcutorture.fwd_progress_holdoff= [KNL]
4776 Number of seconds to wait between successive
4777 forward-progress tests.
4779 rcutorture.fwd_progress_need_resched= [KNL]
4780 Enclose cond_resched() calls within checks for
4781 need_resched() during tight-loop forward-progress
4784 rcutorture.gp_cond= [KNL]
4785 Use conditional/asynchronous update-side
4786 primitives, if available.
4788 rcutorture.gp_exp= [KNL]
4789 Use expedited update-side primitives, if available.
4791 rcutorture.gp_normal= [KNL]
4792 Use normal (non-expedited) asynchronous
4793 update-side primitives, if available.
4795 rcutorture.gp_sync= [KNL]
4796 Use normal (non-expedited) synchronous
4797 update-side primitives, if available. If all
4798 of rcutorture.gp_cond=, rcutorture.gp_exp=,
4799 rcutorture.gp_normal=, and rcutorture.gp_sync=
4800 are zero, rcutorture acts as if is interpreted
4801 they are all non-zero.
4803 rcutorture.irqreader= [KNL]
4804 Run RCU readers from irq handlers, or, more
4805 accurately, from a timer handler. Not all RCU
4806 flavors take kindly to this sort of thing.
4808 rcutorture.leakpointer= [KNL]
4809 Leak an RCU-protected pointer out of the reader.
4810 This can of course result in splats, and is
4811 intended to test the ability of things like
4812 CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect
4815 rcutorture.n_barrier_cbs= [KNL]
4816 Set callbacks/threads for rcu_barrier() testing.
4818 rcutorture.nfakewriters= [KNL]
4819 Set number of concurrent RCU writers. These just
4820 stress RCU, they don't participate in the actual
4821 test, hence the "fake".
4823 rcutorture.nocbs_nthreads= [KNL]
4824 Set number of RCU callback-offload togglers.
4825 Zero (the default) disables toggling.
4827 rcutorture.nocbs_toggle= [KNL]
4828 Set the delay in milliseconds between successive
4829 callback-offload toggling attempts.
4831 rcutorture.nreaders= [KNL]
4832 Set number of RCU readers. The value -1 selects
4833 N-1, where N is the number of CPUs. A value
4834 "n" less than -1 selects N-n-2, where N is again
4835 the number of CPUs. For example, -2 selects N
4836 (the number of CPUs), -3 selects N+1, and so on.
4838 rcutorture.object_debug= [KNL]
4839 Enable debug-object double-call_rcu() testing.
4841 rcutorture.onoff_holdoff= [KNL]
4842 Set time (s) after boot for CPU-hotplug testing.
4844 rcutorture.onoff_interval= [KNL]
4845 Set time (jiffies) between CPU-hotplug operations,
4846 or zero to disable CPU-hotplug testing.
4848 rcutorture.read_exit= [KNL]
4849 Set the number of read-then-exit kthreads used
4850 to test the interaction of RCU updaters and
4851 task-exit processing.
4853 rcutorture.read_exit_burst= [KNL]
4854 The number of times in a given read-then-exit
4855 episode that a set of read-then-exit kthreads
4858 rcutorture.read_exit_delay= [KNL]
4859 The delay, in seconds, between successive
4860 read-then-exit testing episodes.
4862 rcutorture.shuffle_interval= [KNL]
4863 Set task-shuffle interval (s). Shuffling tasks
4864 allows some CPUs to go into dyntick-idle mode
4865 during the rcutorture test.
4867 rcutorture.shutdown_secs= [KNL]
4868 Set time (s) after boot system shutdown. This
4869 is useful for hands-off automated testing.
4871 rcutorture.stall_cpu= [KNL]
4872 Duration of CPU stall (s) to test RCU CPU stall
4873 warnings, zero to disable.
4875 rcutorture.stall_cpu_block= [KNL]
4876 Sleep while stalling if set. This will result
4877 in warnings from preemptible RCU in addition
4878 to any other stall-related activity.
4880 rcutorture.stall_cpu_holdoff= [KNL]
4881 Time to wait (s) after boot before inducing stall.
4883 rcutorture.stall_cpu_irqsoff= [KNL]
4884 Disable interrupts while stalling if set.
4886 rcutorture.stall_gp_kthread= [KNL]
4887 Duration (s) of forced sleep within RCU
4888 grace-period kthread to test RCU CPU stall
4889 warnings, zero to disable. If both stall_cpu
4890 and stall_gp_kthread are specified, the
4891 kthread is starved first, then the CPU.
4893 rcutorture.stat_interval= [KNL]
4894 Time (s) between statistics printk()s.
4896 rcutorture.stutter= [KNL]
4897 Time (s) to stutter testing, for example, specifying
4898 five seconds causes the test to run for five seconds,
4899 wait for five seconds, and so on. This tests RCU's
4900 ability to transition abruptly to and from idle.
4902 rcutorture.test_boost= [KNL]
4903 Test RCU priority boosting? 0=no, 1=maybe, 2=yes.
4904 "Maybe" means test if the RCU implementation
4905 under test support RCU priority boosting.
4907 rcutorture.test_boost_duration= [KNL]
4908 Duration (s) of each individual boost test.
4910 rcutorture.test_boost_interval= [KNL]
4911 Interval (s) between each boost test.
4913 rcutorture.test_no_idle_hz= [KNL]
4914 Test RCU's dyntick-idle handling. See also the
4915 rcutorture.shuffle_interval parameter.
4917 rcutorture.torture_type= [KNL]
4918 Specify the RCU implementation to test.
4920 rcutorture.verbose= [KNL]
4921 Enable additional printk() statements.
4923 rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
4924 Dump ftrace buffer after reporting RCU CPU
4927 rcupdate.rcu_cpu_stall_suppress= [KNL]
4928 Suppress RCU CPU stall warning messages.
4930 rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL]
4931 Suppress RCU CPU stall warning messages and
4932 rcutorture writer stall warnings that occur
4933 during early boot, that is, during the time
4934 before the init task is spawned.
4936 rcupdate.rcu_cpu_stall_timeout= [KNL]
4937 Set timeout for RCU CPU stall warning messages.
4938 The value is in seconds and the maximum allowed
4939 value is 300 seconds.
4941 rcupdate.rcu_exp_cpu_stall_timeout= [KNL]
4942 Set timeout for expedited RCU CPU stall warning
4943 messages. The value is in milliseconds
4944 and the maximum allowed value is 21000
4945 milliseconds. Please note that this value is
4946 adjusted to an arch timer tick resolution.
4947 Setting this to zero causes the value from
4948 rcupdate.rcu_cpu_stall_timeout to be used (after
4949 conversion from seconds to milliseconds).
4951 rcupdate.rcu_expedited= [KNL]
4952 Use expedited grace-period primitives, for
4953 example, synchronize_rcu_expedited() instead
4954 of synchronize_rcu(). This reduces latency,
4955 but can increase CPU utilization, degrade
4956 real-time latency, and degrade energy efficiency.
4957 No effect on CONFIG_TINY_RCU kernels.
4959 rcupdate.rcu_normal= [KNL]
4960 Use only normal grace-period primitives,
4961 for example, synchronize_rcu() instead of
4962 synchronize_rcu_expedited(). This improves
4963 real-time latency, CPU utilization, and
4964 energy efficiency, but can expose users to
4965 increased grace-period latency. This parameter
4966 overrides rcupdate.rcu_expedited. No effect on
4967 CONFIG_TINY_RCU kernels.
4969 rcupdate.rcu_normal_after_boot= [KNL]
4970 Once boot has completed (that is, after
4971 rcu_end_inkernel_boot() has been invoked), use
4972 only normal grace-period primitives. No effect
4973 on CONFIG_TINY_RCU kernels.
4975 But note that CONFIG_PREEMPT_RT=y kernels enables
4976 this kernel boot parameter, forcibly setting
4977 it to the value one, that is, converting any
4978 post-boot attempt at an expedited RCU grace
4979 period to instead use normal non-expedited
4980 grace-period processing.
4982 rcupdate.rcu_task_collapse_lim= [KNL]
4983 Set the maximum number of callbacks present
4984 at the beginning of a grace period that allows
4985 the RCU Tasks flavors to collapse back to using
4986 a single callback queue. This switching only
4987 occurs when rcupdate.rcu_task_enqueue_lim is
4988 set to the default value of -1.
4990 rcupdate.rcu_task_contend_lim= [KNL]
4991 Set the minimum number of callback-queuing-time
4992 lock-contention events per jiffy required to
4993 cause the RCU Tasks flavors to switch to per-CPU
4994 callback queuing. This switching only occurs
4995 when rcupdate.rcu_task_enqueue_lim is set to
4996 the default value of -1.
4998 rcupdate.rcu_task_enqueue_lim= [KNL]
4999 Set the number of callback queues to use for the
5000 RCU Tasks family of RCU flavors. The default
5001 of -1 allows this to be automatically (and
5002 dynamically) adjusted. This parameter is intended
5005 rcupdate.rcu_task_ipi_delay= [KNL]
5006 Set time in jiffies during which RCU tasks will
5007 avoid sending IPIs, starting with the beginning
5008 of a given grace period. Setting a large
5009 number avoids disturbing real-time workloads,
5010 but lengthens grace periods.
5012 rcupdate.rcu_task_stall_info= [KNL]
5013 Set initial timeout in jiffies for RCU task stall
5014 informational messages, which give some indication
5015 of the problem for those not patient enough to
5016 wait for ten minutes. Informational messages are
5017 only printed prior to the stall-warning message
5018 for a given grace period. Disable with a value
5019 less than or equal to zero. Defaults to ten
5020 seconds. A change in value does not take effect
5021 until the beginning of the next grace period.
5023 rcupdate.rcu_task_stall_info_mult= [KNL]
5024 Multiplier for time interval between successive
5025 RCU task stall informational messages for a given
5026 RCU tasks grace period. This value is clamped
5027 to one through ten, inclusive. It defaults to
5028 the value three, so that the first informational
5029 message is printed 10 seconds into the grace
5030 period, the second at 40 seconds, the third at
5031 160 seconds, and then the stall warning at 600
5032 seconds would prevent a fourth at 640 seconds.
5034 rcupdate.rcu_task_stall_timeout= [KNL]
5035 Set timeout in jiffies for RCU task stall
5036 warning messages. Disable with a value less
5037 than or equal to zero. Defaults to ten minutes.
5038 A change in value does not take effect until
5039 the beginning of the next grace period.
5041 rcupdate.rcu_self_test= [KNL]
5042 Run the RCU early boot self tests
5046 Run specified binary instead of /init from the ramdisk,
5047 used for early userspace startup. See initrd.
5050 force - Override the decision by the kernel to hide the
5051 advertisement of RDRAND support (this affects
5052 certain AMD processors because of buggy BIOS
5053 support, specifically around the suspend/resume
5057 Turn on/off individual RDT features. List is:
5058 cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp,
5060 E.g. to turn on cmt and turn off mba use:
5064 Format (x86 or x86_64):
5065 [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \
5067 [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \
5069 Where reboot_mode is one of warm (soft) or cold (hard) or gpio
5070 (prefix with 'panic_' to set mode for panic
5072 reboot_type is one of bios, acpi, kbd, triple, efi, or pci,
5073 reboot_force is either force or not specified,
5074 reboot_cpu is s[mp]#### with #### being the processor
5075 to be used for rebooting.
5077 refscale.holdoff= [KNL]
5078 Set test-start holdoff period. The purpose of
5079 this parameter is to delay the start of the
5080 test until boot completes in order to avoid
5083 refscale.loops= [KNL]
5084 Set the number of loops over the synchronization
5085 primitive under test. Increasing this number
5086 reduces noise due to loop start/end overhead,
5087 but the default has already reduced the per-pass
5088 noise to a handful of picoseconds on ca. 2020
5091 refscale.nreaders= [KNL]
5092 Set number of readers. The default value of -1
5093 selects N, where N is roughly 75% of the number
5094 of CPUs. A value of zero is an interesting choice.
5096 refscale.nruns= [KNL]
5097 Set number of runs, each of which is dumped onto
5100 refscale.readdelay= [KNL]
5101 Set the read-side critical-section duration,
5102 measured in microseconds.
5104 refscale.scale_type= [KNL]
5105 Specify the read-protection implementation to test.
5107 refscale.shutdown= [KNL]
5108 Shut down the system at the end of the performance
5109 test. This defaults to 1 (shut it down) when
5110 refscale is built into the kernel and to 0 (leave
5111 it running) when refscale is built as a module.
5113 refscale.verbose= [KNL]
5114 Enable additional printk() statements.
5116 refscale.verbose_batched= [KNL]
5117 Batch the additional printk() statements. If zero
5118 (the default) or negative, print everything. Otherwise,
5119 print every Nth verbose statement, where N is the value
5123 [KNL, SMP] Set scheduler's default relax_domain_level.
5124 See Documentation/admin-guide/cgroup-v1/cpusets.rst.
5126 reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory
5127 Format: <base1>,<size1>[,<base2>,<size2>,...]
5128 Reserve I/O ports or memory so the kernel won't use
5129 them. If <base> is less than 0x10000, the region
5130 is assumed to be I/O ports; otherwise it is memory.
5132 reservetop= [X86-32]
5134 Reserves a hole at the top of the kernel virtual
5137 reset_devices [KNL] Force drivers to reset the underlying device
5138 during initialization.
5141 Specify the partition device for software suspend
5143 {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>}
5145 resume_offset= [SWSUSP]
5146 Specify the offset from the beginning of the partition
5147 given by "resume=" at which the swap header is located,
5148 in <PAGE_SIZE> units (needed only for swap files).
5149 See Documentation/power/swsusp-and-swap-files.rst
5151 resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to
5152 read the resume files
5154 resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up.
5155 Useful for devices that are detected asynchronously
5156 (e.g. USB and MMC devices).
5158 retain_initrd [RAM] Keep initrd memory after extraction
5160 rfkill.default_state=
5161 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm,
5162 etc. communication is blocked by default.
5165 rfkill.master_switch_mode=
5166 0 The "airplane mode" button does nothing.
5167 1 The "airplane mode" button toggles between everything
5168 blocked and the previous configuration.
5169 2 The "airplane mode" button toggles between everything
5170 blocked and everything unblocked.
5172 rhash_entries= [KNL,NET]
5173 Set number of hash buckets for route cache
5176 [KNL] Disable ring 3 MONITOR/MWAIT feature on supported
5179 ro [KNL] Mount root device read-only on boot
5182 on Mark read-only kernel memory as read-only (default).
5183 off Leave read-only kernel memory writable for debugging.
5186 Enable the uart passthrough on the designated usb port
5187 on Rockchip SoCs. When active, the signals of the
5188 debug-uart get routed to the D+ and D- pins of the usb
5189 port and the regular usb controller gets disabled.
5191 root= [KNL] Root filesystem
5192 See name_to_dev_t comment in init/do_mounts.c.
5194 rootdelay= [KNL] Delay (in seconds) to pause before attempting to
5195 mount the root filesystem
5197 rootflags= [KNL] Set root filesystem mount option string
5199 rootfstype= [KNL] Set root filesystem type
5201 rootwait [KNL] Wait (indefinitely) for root device to show up.
5202 Useful for devices that are detected asynchronously
5203 (e.g. USB and MMC devices).
5205 rproc_mem=nn[KMG][@address]
5206 [KNL,ARM,CMA] Remoteproc physical memory block.
5207 Memory area to be used by remote processor image,
5210 rw [KNL] Mount root device read-write on boot
5212 S [KNL] Run init in single mode
5214 s390_iommu= [HW,S390]
5215 Set s390 IOTLB flushing mode
5217 With strict flushing every unmap operation will result in
5218 an IOTLB flush. Default is lazy flushing before reuse,
5221 s390_iommu_aperture= [KNL,S390]
5222 Specifies the size of the per device DMA address space
5223 accessible through the DMA and IOMMU APIs as a decimal
5224 factor of the size of main memory.
5225 The default is 1 meaning that one can concurrently use
5226 as many DMA addresses as physical memory is installed,
5227 if supported by hardware, and thus map all of memory
5228 once. With a value of 2 one can map all of memory twice
5229 and so on. As a special case a factor of 0 imposes no
5230 restrictions other than those given by hardware at the
5231 cost of significant additional memory use for tables.
5234 See drivers/net/irda/sa1100_ir.c.
5236 sched_verbose [KNL] Enables verbose scheduler debug messages.
5238 schedstats= [KNL,X86] Enable or disable scheduled statistics.
5239 Allowed values are enable and disable. This feature
5240 incurs a small amount of overhead in the scheduler
5241 but is useful for debugging and performance tuning.
5243 sched_thermal_decay_shift=
5244 [KNL, SMP] Set a decay shift for scheduler thermal
5245 pressure signal. Thermal pressure signal follows the
5246 default decay period of other scheduler pelt
5247 signals(usually 32 ms but configurable). Setting
5248 sched_thermal_decay_shift will left shift the decay
5249 period for the thermal pressure signal by the shift
5251 i.e. with the default pelt decay period of 32 ms
5252 sched_thermal_decay_shift thermal pressure decay pr
5256 Format: integer between 0 and 10
5259 scftorture.holdoff= [KNL]
5260 Number of seconds to hold off before starting
5261 test. Defaults to zero for module insertion and
5262 to 10 seconds for built-in smp_call_function()
5265 scftorture.longwait= [KNL]
5266 Request ridiculously long waits randomly selected
5267 up to the chosen limit in seconds. Zero (the
5268 default) disables this feature. Please note
5269 that requesting even small non-zero numbers of
5270 seconds can result in RCU CPU stall warnings,
5271 softlockup complaints, and so on.
5273 scftorture.nthreads= [KNL]
5274 Number of kthreads to spawn to invoke the
5275 smp_call_function() family of functions.
5276 The default of -1 specifies a number of kthreads
5277 equal to the number of CPUs.
5279 scftorture.onoff_holdoff= [KNL]
5280 Number seconds to wait after the start of the
5281 test before initiating CPU-hotplug operations.
5283 scftorture.onoff_interval= [KNL]
5284 Number seconds to wait between successive
5285 CPU-hotplug operations. Specifying zero (which
5286 is the default) disables CPU-hotplug operations.
5288 scftorture.shutdown_secs= [KNL]
5289 The number of seconds following the start of the
5290 test after which to shut down the system. The
5291 default of zero avoids shutting down the system.
5292 Non-zero values are useful for automated tests.
5294 scftorture.stat_interval= [KNL]
5295 The number of seconds between outputting the
5296 current test statistics to the console. A value
5297 of zero disables statistics output.
5299 scftorture.stutter_cpus= [KNL]
5300 The number of jiffies to wait between each change
5301 to the set of CPUs under test.
5303 scftorture.use_cpus_read_lock= [KNL]
5304 Use use_cpus_read_lock() instead of the default
5305 preempt_disable() to disable CPU hotplug
5306 while invoking one of the smp_call_function*()
5309 scftorture.verbose= [KNL]
5310 Enable additional printk() statements.
5312 scftorture.weight_single= [KNL]
5313 The probability weighting to use for the
5314 smp_call_function_single() function with a zero
5315 "wait" parameter. A value of -1 selects the
5316 default if all other weights are -1. However,
5317 if at least one weight has some other value, a
5318 value of -1 will instead select a weight of zero.
5320 scftorture.weight_single_wait= [KNL]
5321 The probability weighting to use for the
5322 smp_call_function_single() function with a
5323 non-zero "wait" parameter. See weight_single.
5325 scftorture.weight_many= [KNL]
5326 The probability weighting to use for the
5327 smp_call_function_many() function with a zero
5328 "wait" parameter. See weight_single.
5329 Note well that setting a high probability for
5330 this weighting can place serious IPI load
5333 scftorture.weight_many_wait= [KNL]
5334 The probability weighting to use for the
5335 smp_call_function_many() function with a
5336 non-zero "wait" parameter. See weight_single
5339 scftorture.weight_all= [KNL]
5340 The probability weighting to use for the
5341 smp_call_function_all() function with a zero
5342 "wait" parameter. See weight_single and
5345 scftorture.weight_all_wait= [KNL]
5346 The probability weighting to use for the
5347 smp_call_function_all() function with a
5348 non-zero "wait" parameter. See weight_single
5351 skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate
5352 xtime_lock contention on larger systems, and/or RCU lock
5353 contention on all systems with CONFIG_MAXSMP set.
5354 Format: { "0" | "1" }
5355 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1"
5357 Note: increases power consumption, thus should only be
5358 enabled if running jitter sensitive (HPC/RT) workloads.
5360 security= [SECURITY] Choose a legacy "major" security module to
5361 enable at boot. This has been deprecated by the
5364 selinux= [SELINUX] Disable or enable SELinux at boot time.
5365 Format: { "0" | "1" }
5366 See security/selinux/Kconfig help text.
5371 apparmor= [APPARMOR] Disable or enable AppArmor at boot time
5372 Format: { "0" | "1" }
5373 See security/apparmor/Kconfig help text
5376 Default value is set via kernel config option.
5378 serialnumber [BUGS=X86-32]
5380 sev=option[,option...] [X86-64] See Documentation/x86/x86_64/boot-options.rst
5383 Maximal number of shapers.
5391 Enable merging of slabs with similar size when the
5392 kernel is built without CONFIG_SLAB_MERGE_DEFAULT.
5395 Disable merging of slabs with similar size. May be
5396 necessary if there is some reason to distinguish
5397 allocs to different slabs, especially in hardened
5398 environments where the risk of heap overflows and
5399 layout control by attackers can usually be
5400 frustrated by disabling merging. This will reduce
5401 most of the exposure of a heap attack to a single
5402 cache (risks via metadata attacks are mostly
5403 unchanged). Debug options disable merging on their
5405 For more information see Documentation/vm/slub.rst.
5407 slab_max_order= [MM, SLAB]
5408 Determines the maximum allowed order for slabs.
5409 A high setting may cause OOMs due to memory
5410 fragmentation. Defaults to 1 for systems with
5411 more than 32MB of RAM, 0 otherwise.
5413 slub_debug[=options[,slabs][;[options[,slabs]]...] [MM, SLUB]
5414 Enabling slub_debug allows one to determine the
5415 culprit if slab objects become corrupted. Enabling
5416 slub_debug can create guard zones around objects and
5417 may poison objects when not in use. Also tracks the
5418 last alloc / free. For more information see
5419 Documentation/vm/slub.rst.
5421 slub_max_order= [MM, SLUB]
5422 Determines the maximum allowed order for slabs.
5423 A high setting may cause OOMs due to memory
5424 fragmentation. For more information see
5425 Documentation/vm/slub.rst.
5427 slub_min_objects= [MM, SLUB]
5428 The minimum number of objects per slab. SLUB will
5429 increase the slab order up to slub_max_order to
5430 generate a sufficiently large slab able to contain
5431 the number of objects indicated. The higher the number
5432 of objects the smaller the overhead of tracking slabs
5433 and the less frequently locks need to be acquired.
5434 For more information see Documentation/vm/slub.rst.
5436 slub_min_order= [MM, SLUB]
5437 Determines the minimum page order for slabs. Must be
5438 lower than slub_max_order.
5439 For more information see Documentation/vm/slub.rst.
5441 slub_merge [MM, SLUB]
5442 Same with slab_merge.
5444 slub_nomerge [MM, SLUB]
5445 Same with slab_nomerge. This is supported for legacy.
5446 See slab_nomerge for more information.
5449 Format: <io1>[,<io2>[,...,<io8>]]
5451 smp.csd_lock_timeout= [KNL]
5452 Specify the period of time in milliseconds
5453 that smp_call_function() and friends will wait
5454 for a CPU to release the CSD lock. This is
5455 useful when diagnosing bugs involving CPUs
5456 disabling interrupts for extended periods
5457 of time. Defaults to 5,000 milliseconds, and
5458 setting a value of zero disables this feature.
5459 This feature may be more efficiently disabled
5460 using the csdlock_debug- kernel parameter.
5462 smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices
5463 smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port
5464 smsc-ircc2.ircc_sir= [HW] SIR base I/O port
5465 smsc-ircc2.ircc_fir= [HW] FIR base I/O port
5466 smsc-ircc2.ircc_irq= [HW] IRQ line
5467 smsc-ircc2.ircc_dma= [HW] DMA channel
5468 smsc-ircc2.ircc_transceiver= [HW] Transceiver type:
5469 0: Toshiba Satellite 1800 (GP data pin select)
5470 1: Fast pin select (default)
5473 smt= [KNL,S390] Set the maximum number of threads (logical
5474 CPUs) to use per physical CPU on systems capable of
5475 symmetric multithreading (SMT). Will be capped to the
5476 actual hardware limit.
5478 Default: -1 (no limit)
5481 [KNL] Should the soft-lockup detector generate panics.
5484 A value of 1 instructs the soft-lockup detector
5485 to panic the machine when a soft-lockup occurs. It is
5486 also controlled by the kernel.softlockup_panic sysctl
5487 and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the
5488 respective build-time switch to that functionality.
5490 softlockup_all_cpu_backtrace=
5491 [KNL] Should the soft-lockup detector generate
5492 backtraces on all cpus.
5495 sonypi.*= [HW] Sony Programmable I/O Control Device driver
5496 See Documentation/admin-guide/laptops/sonypi.rst
5498 spectre_v2= [X86] Control mitigation of Spectre variant 2
5499 (indirect branch speculation) vulnerability.
5500 The default operation protects the kernel from
5503 on - unconditionally enable, implies
5505 off - unconditionally disable, implies
5507 auto - kernel detects whether your CPU model is
5510 Selecting 'on' will, and 'auto' may, choose a
5511 mitigation method at run time according to the
5512 CPU, the available microcode, the setting of the
5513 CONFIG_RETPOLINE configuration option, and the
5514 compiler with which the kernel was built.
5516 Selecting 'on' will also enable the mitigation
5517 against user space to user space task attacks.
5519 Selecting 'off' will disable both the kernel and
5520 the user space protections.
5522 Specific mitigations can also be selected manually:
5524 retpoline - replace indirect branches
5525 retpoline,generic - Retpolines
5526 retpoline,lfence - LFENCE; indirect branch
5527 retpoline,amd - alias for retpoline,lfence
5528 eibrs - enhanced IBRS
5529 eibrs,retpoline - enhanced IBRS + Retpolines
5530 eibrs,lfence - enhanced IBRS + LFENCE
5532 Not specifying this option is equivalent to
5536 [X86] Control mitigation of Spectre variant 2
5537 (indirect branch speculation) vulnerability between
5540 on - Unconditionally enable mitigations. Is
5541 enforced by spectre_v2=on
5543 off - Unconditionally disable mitigations. Is
5544 enforced by spectre_v2=off
5546 prctl - Indirect branch speculation is enabled,
5547 but mitigation can be enabled via prctl
5548 per thread. The mitigation control state
5549 is inherited on fork.
5552 - Like "prctl" above, but only STIBP is
5553 controlled per thread. IBPB is issued
5554 always when switching between different user
5558 - Same as "prctl" above, but all seccomp
5559 threads will enable the mitigation unless
5560 they explicitly opt out.
5563 - Like "seccomp" above, but only STIBP is
5564 controlled per thread. IBPB is issued
5565 always when switching between different
5566 user space processes.
5568 auto - Kernel selects the mitigation depending on
5569 the available CPU features and vulnerability.
5571 Default mitigation: "prctl"
5573 Not specifying this option is equivalent to
5574 spectre_v2_user=auto.
5576 spec_store_bypass_disable=
5577 [HW] Control Speculative Store Bypass (SSB) Disable mitigation
5578 (Speculative Store Bypass vulnerability)
5580 Certain CPUs are vulnerable to an exploit against a
5581 a common industry wide performance optimization known
5582 as "Speculative Store Bypass" in which recent stores
5583 to the same memory location may not be observed by
5584 later loads during speculative execution. The idea
5585 is that such stores are unlikely and that they can
5586 be detected prior to instruction retirement at the
5587 end of a particular speculation execution window.
5589 In vulnerable processors, the speculatively forwarded
5590 store can be used in a cache side channel attack, for
5591 example to read memory to which the attacker does not
5592 directly have access (e.g. inside sandboxed code).
5594 This parameter controls whether the Speculative Store
5595 Bypass optimization is used.
5597 On x86 the options are:
5599 on - Unconditionally disable Speculative Store Bypass
5600 off - Unconditionally enable Speculative Store Bypass
5601 auto - Kernel detects whether the CPU model contains an
5602 implementation of Speculative Store Bypass and
5603 picks the most appropriate mitigation. If the
5604 CPU is not vulnerable, "off" is selected. If the
5605 CPU is vulnerable the default mitigation is
5606 architecture and Kconfig dependent. See below.
5607 prctl - Control Speculative Store Bypass per thread
5608 via prctl. Speculative Store Bypass is enabled
5609 for a process by default. The state of the control
5610 is inherited on fork.
5611 seccomp - Same as "prctl" above, but all seccomp threads
5612 will disable SSB unless they explicitly opt out.
5614 Default mitigations:
5617 On powerpc the options are:
5619 on,auto - On Power8 and Power9 insert a store-forwarding
5620 barrier on kernel entry and exit. On Power7
5621 perform a software flush on kernel entry and
5625 Not specifying this option is equivalent to
5626 spec_store_bypass_disable=auto.
5628 spia_io_base= [HW,MTD]
5634 [X86] Enable split lock detection or bus lock detection
5636 When enabled (and if hardware support is present), atomic
5637 instructions that access data across cache line
5638 boundaries will result in an alignment check exception
5639 for split lock detection or a debug exception for
5644 warn - the kernel will emit rate-limited warnings
5645 about applications triggering the #AC
5646 exception or the #DB exception. This mode is
5647 the default on CPUs that support split lock
5648 detection or bus lock detection. Default
5649 behavior is by #AC if both features are
5650 enabled in hardware.
5652 fatal - the kernel will send SIGBUS to applications
5653 that trigger the #AC exception or the #DB
5654 exception. Default behavior is by #AC if
5655 both features are enabled in hardware.
5658 Set system wide rate limit to N bus locks
5659 per second for bus lock detection.
5662 N/A for split lock detection.
5665 If an #AC exception is hit in the kernel or in
5666 firmware (i.e. not while executing in user mode)
5667 the kernel will oops in either "warn" or "fatal"
5670 #DB exception for bus lock is triggered only when
5674 Control the Special Register Buffer Data Sampling
5677 Certain CPUs are vulnerable to an MDS-like
5678 exploit which can leak bits from the random
5681 By default, this issue is mitigated by
5682 microcode. However, the microcode fix can cause
5683 the RDRAND and RDSEED instructions to become
5684 much slower. Among other effects, this will
5685 result in reduced throughput from /dev/urandom.
5687 The microcode mitigation can be disabled with
5688 the following option:
5690 off: Disable mitigation and remove
5691 performance impact to RDRAND and RDSEED
5693 srcutree.big_cpu_lim [KNL]
5694 Specifies the number of CPUs constituting a
5695 large system, such that srcu_struct structures
5696 should immediately allocate an srcu_node array.
5697 This kernel-boot parameter defaults to 128,
5698 but takes effect only when the low-order four
5699 bits of srcutree.convert_to_big is equal to 3
5702 srcutree.convert_to_big [KNL]
5703 Specifies under what conditions an SRCU tree
5704 srcu_struct structure will be converted to big
5705 form, that is, with an rcu_node tree:
5708 1: At init_srcu_struct() time.
5709 2: When rcutorture decides to.
5710 3: Decide at boot time (default).
5711 0x1X: Above plus if high contention.
5713 Either way, the srcu_node tree will be sized based
5714 on the actual runtime number of CPUs (nr_cpu_ids)
5715 instead of the compile-time CONFIG_NR_CPUS.
5717 srcutree.counter_wrap_check [KNL]
5718 Specifies how frequently to check for
5719 grace-period sequence counter wrap for the
5720 srcu_data structure's ->srcu_gp_seq_needed field.
5721 The greater the number of bits set in this kernel
5722 parameter, the less frequently counter wrap will
5723 be checked for. Note that the bottom two bits
5726 srcutree.exp_holdoff [KNL]
5727 Specifies how many nanoseconds must elapse
5728 since the end of the last SRCU grace period for
5729 a given srcu_struct until the next normal SRCU
5730 grace period will be considered for automatic
5731 expediting. Set to zero to disable automatic
5734 srcutree.small_contention_lim [KNL]
5735 Specifies the number of update-side contention
5736 events per jiffy will be tolerated before
5737 initiating a conversion of an srcu_struct
5738 structure to big form. Note that the value of
5739 srcutree.convert_to_big must have the 0x10 bit
5740 set for contention-based conversions to occur.
5743 Speculative Store Bypass Disable control
5745 On CPUs that are vulnerable to the Speculative
5746 Store Bypass vulnerability and offer a
5747 firmware based mitigation, this parameter
5748 indicates how the mitigation should be used:
5750 force-on: Unconditionally enable mitigation for
5751 for both kernel and userspace
5752 force-off: Unconditionally disable mitigation for
5753 for both kernel and userspace
5754 kernel: Always enable mitigation in the
5755 kernel, and offer a prctl interface
5756 to allow userspace to register its
5757 interest in being mitigated too.
5759 stack_guard_gap= [MM]
5760 override the default stack gap protection. The value
5761 is in page units and it defines how many pages prior
5762 to (for stacks growing down) resp. after (for stacks
5763 growing up) the main stack are reserved for no other
5764 mapping. Default value is 256 pages.
5766 stack_depot_disable= [KNL]
5767 Setting this to true through kernel command line will
5768 disable the stack depot thereby saving the static memory
5769 consumed by the stack hash table. By default this is set
5773 Enabled the stack tracer on boot up.
5775 stacktrace_filter=[function-list]
5776 [FTRACE] Limit the functions that the stack tracer
5777 will trace at boot up. function-list is a comma-separated
5778 list of functions. This list can be changed at run
5779 time by the stack_trace_filter file in the debugfs
5780 tracing directory. Note, this enables stack tracing
5781 and the stacktrace above is not needed.
5785 Set the STI (builtin display/keyboard on the HP-PARISC
5786 machines) console (graphic card) which should be used
5787 as the initial boot-console.
5788 See also comment in drivers/video/console/sticore.c.
5791 See comment in drivers/video/console/sticore.c.
5794 Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]]
5799 Enable or disable strict sigaltstack size checks
5800 against the required signal frame size which
5801 depends on the supported FPU features. This can
5802 be used to filter out binaries which have
5803 not yet been made aware of AT_MINSIGSTKSZ.
5805 sunrpc.min_resvport=
5806 sunrpc.max_resvport=
5808 SunRPC servers often require that client requests
5809 originate from a privileged port (i.e. a port in the
5810 range 0 < portnr < 1024).
5811 An administrator who wishes to reserve some of these
5812 ports for other uses may adjust the range that the
5813 kernel's sunrpc client considers to be privileged
5814 using these two parameters to set the minimum and
5815 maximum port values.
5817 sunrpc.svc_rpc_per_connection_limit=
5819 Limit the number of requests that the server will
5820 process in parallel from a single connection.
5821 The default value is 0 (no limit).
5825 Control how the NFS server code allocates CPUs to
5826 service thread pools. Depending on how many NICs
5827 you have and where their interrupts are bound, this
5828 option will affect which CPUs will do NFS serving.
5829 Note: this parameter cannot be changed while the
5830 NFS server is running.
5832 auto the server chooses an appropriate mode
5833 automatically using heuristics
5834 global a single global pool contains all CPUs
5835 percpu one pool for each CPU
5836 pernode one pool for each NUMA node (equivalent
5837 to global on non-NUMA machines)
5839 sunrpc.tcp_slot_table_entries=
5840 sunrpc.udp_slot_table_entries=
5842 Sets the upper limit on the number of simultaneous
5843 RPC calls that can be sent from the client to a
5844 server. Increasing these values may allow you to
5845 improve throughput, but will also increase the
5846 amount of memory reserved for use by the client.
5848 suspend.pm_test_delay=
5850 Sets the number of seconds to remain in a suspend test
5851 mode before resuming the system (see
5852 /sys/power/pm_test). Only available when CONFIG_PM_DEBUG
5853 is set. Default value is 5.
5856 Format: { on | off | y | n | 1 | 0 }
5857 This parameter controls use of the Protected
5858 Execution Facility on pSeries.
5862 Enable accounting of swap in memory resource
5863 controller if no parameter or 1 is given or disable
5864 it if 0 is given (See Documentation/admin-guide/cgroup-v1/memory.rst)
5866 swiotlb= [ARM,IA-64,PPC,MIPS,X86]
5867 Format: { <int> | force | noforce }
5868 <int> -- Number of I/O TLB slabs
5869 force -- force using of bounce buffers even if they
5870 wouldn't be automatically used by the kernel
5871 noforce -- Never use bounce buffers (for debugging)
5876 Set a sysctl parameter, right before loading the init
5877 process, as if the value was written to the respective
5878 /proc/sys/... file. Both '.' and '/' are recognized as
5879 separators. Unrecognized parameters and invalid values
5880 are reported in the kernel log. Sysctls registered
5881 later by a loaded module cannot be set this way.
5882 Example: sysctl.vm.swappiness=40
5884 sysfs.deprecated=0|1 [KNL]
5885 Enable/disable old style sysfs layout for old udev
5886 on older distributions. When this option is enabled
5887 very new udev will not work anymore. When this option
5888 is disabled (or CONFIG_SYSFS_DEPRECATED not compiled)
5889 in older udev will not work anymore.
5890 Default depends on CONFIG_SYSFS_DEPRECATED_V2 set in
5891 the kernel configuration.
5893 sysrq_always_enabled
5895 Ignore sysrq setting - this boot parameter will
5896 neutralize any effect of /proc/sys/kernel/sysrq.
5897 Useful for debugging.
5899 tcpmhash_entries= [KNL,NET]
5900 Set the number of tcp_metrics_hash slots.
5901 Default value is 8192 or 16384 depending on total
5902 ram pages. This is used to specify the TCP metrics
5903 cache size. See Documentation/networking/ip-sysctl.rst
5904 "tcp_no_metrics_save" section for more details.
5908 test_suspend= [SUSPEND]
5909 Format: { "mem" | "standby" | "freeze" }[,N]
5910 Specify "mem" (for Suspend-to-RAM) or "standby" (for
5911 standby suspend) or "freeze" (for suspend type freeze)
5912 as the system sleep state during system startup with
5913 the optional capability to repeat N number of times.
5914 The system is woken from this state using a
5915 wakeup-capable RTC alarm.
5917 thash_entries= [KNL,NET]
5918 Set number of hash buckets for TCP connection
5920 thermal.act= [HW,ACPI]
5921 -1: disable all active trip points in all thermal zones
5922 <degrees C>: override all lowest active trip points
5924 thermal.crt= [HW,ACPI]
5925 -1: disable all critical trip points in all thermal zones
5926 <degrees C>: override all critical trip points
5928 thermal.nocrt= [HW,ACPI]
5929 Set to disable actions on ACPI thermal zone
5930 critical and hot trip points.
5932 thermal.off= [HW,ACPI]
5933 1: disable ACPI thermal control
5935 thermal.psv= [HW,ACPI]
5936 -1: disable all passive trip points
5937 <degrees C>: override all passive trip points to this
5940 thermal.tzp= [HW,ACPI]
5941 Specify global default ACPI thermal zone polling rate
5942 <deci-seconds>: poll all this frequency
5943 0: no polling (default)
5946 Force threading of all interrupt handlers except those
5947 marked explicitly IRQF_NO_THREAD.
5951 Specify if the kernel should make use of the cpu
5952 topology information if the hardware supports this.
5953 The scheduler will make use of this information and
5954 e.g. base its process migration decisions on it.
5957 topology_updates= [KNL, PPC, NUMA]
5959 Specify if the kernel should ignore (off)
5960 topology updates sent by the hypervisor to this
5963 torture.disable_onoff_at_boot= [KNL]
5964 Prevent the CPU-hotplug component of torturing
5965 until after init has spawned.
5967 torture.ftrace_dump_at_shutdown= [KNL]
5968 Dump the ftrace buffer at torture-test shutdown,
5969 even if there were no errors. This can be a
5970 very costly operation when many torture tests
5971 are running concurrently, especially on systems
5972 with rotating-rust storage.
5974 torture.verbose_sleep_frequency= [KNL]
5975 Specifies how many verbose printk()s should be
5976 emitted between each sleep. The default of zero
5977 disables verbose-printk() sleeping.
5979 torture.verbose_sleep_duration= [KNL]
5980 Duration of each verbose-printk() sleep in jiffies.
5984 tpm_suspend_pcr=[HW,TPM]
5985 Format: integer pcr id
5986 Specify that at suspend time, the tpm driver
5987 should extend the specified pcr with zeros,
5988 as a workaround for some chips which fail to
5989 flush the last written pcr on TPM_SaveState.
5990 This will guarantee that all the other pcrs
5994 Have the tracepoints sent to printk as well as the
5995 tracing ring buffer. This is useful for early boot up
5996 where the system hangs or reboots and does not give the
5997 option for reading the tracing buffer or performing a
5998 ftrace_dump_on_oops.
6000 To turn off having tracepoints sent to printk,
6001 echo 0 > /proc/sys/kernel/tracepoint_printk
6002 Note, echoing 1 into this file without the
6003 tracepoint_printk kernel cmdline option has no effect.
6005 The tp_printk_stop_on_boot (see below) can also be used
6006 to stop the printing of events to console at
6011 Having tracepoints sent to printk() and activating high
6012 frequency tracepoints such as irq or sched, can cause
6013 the system to live lock.
6015 tp_printk_stop_on_boot [FTRACE]
6016 When tp_printk (above) is set, it can cause a lot of noise
6017 on the console. It may be useful to only include the
6018 printing of events during boot up, as user space may
6019 make the system inoperable.
6021 This command line option will stop the printing of events
6022 to console at the late_initcall_sync() time frame.
6024 trace_buf_size=nn[KMG]
6025 [FTRACE] will set tracing buffer size on each cpu.
6027 trace_clock= [FTRACE] Set the clock used for tracing events
6029 local - Use the per CPU time stamp counter
6030 (converted into nanoseconds). Fast, but
6031 depending on the architecture, may not be
6032 in sync between CPUs.
6033 global - Event time stamps are synchronize across
6034 CPUs. May be slower than the local clock,
6035 but better for some race conditions.
6036 counter - Simple counting of events (1, 2, ..)
6037 note, some counts may be skipped due to the
6038 infrastructure grabbing the clock more than
6040 uptime - Use jiffies as the time stamp.
6041 perf - Use the same clock that perf uses.
6042 mono - Use ktime_get_mono_fast_ns() for time stamps.
6043 mono_raw - Use ktime_get_raw_fast_ns() for time
6045 boot - Use ktime_get_boot_fast_ns() for time stamps.
6046 Architectures may add more clocks. See
6047 Documentation/trace/ftrace.rst for more details.
6049 trace_event=[event-list]
6050 [FTRACE] Set and start specified trace events in order
6051 to facilitate early boot debugging. The event-list is a
6052 comma-separated list of trace events to enable. See
6053 also Documentation/trace/events.rst
6055 trace_options=[option-list]
6056 [FTRACE] Enable or disable tracer options at boot.
6057 The option-list is a comma delimited list of options
6058 that can be enabled or disabled just as if you were
6059 to echo the option name into
6061 /sys/kernel/debug/tracing/trace_options
6063 For example, to enable stacktrace option (to dump the
6064 stack trace of each event), add to the command line:
6066 trace_options=stacktrace
6068 See also Documentation/trace/ftrace.rst "trace options"
6072 [FTRACE] enable this option to disable tracing when a
6073 warning is hit. This turns off "tracing_on". Tracing can
6074 be enabled again by echoing '1' into the "tracing_on"
6075 file located in /sys/kernel/debug/tracing/
6077 This option is useful, as it disables the trace before
6078 the WARNING dump is called, which prevents the trace to
6079 be filled with content caused by the warning output.
6081 This option can also be set at run time via the sysctl
6082 option: kernel/traceoff_on_warning
6084 transparent_hugepage=
6086 Format: [always|madvise|never]
6087 Can be used to control the default behavior of the system
6088 with respect to transparent hugepages.
6089 See Documentation/admin-guide/mm/transhuge.rst
6092 trusted.source= [KEYS]
6094 This parameter identifies the trust source as a backend
6095 for trusted keys implementation. Supported trust
6100 If not specified then it defaults to iterating through
6101 the trust source list starting with TPM and assigns the
6102 first trust source as a backend which is initialized
6103 successfully during iteration.
6107 The RNG used to generate key material for trusted keys.
6110 - the same value as trusted.source: "tpm" or "tee"
6112 If not specified, "default" is used. In this case,
6113 the RNG's choice is left to each individual trust source.
6115 tsc= Disable clocksource stability checks for TSC.
6117 [x86] reliable: mark tsc clocksource as reliable, this
6118 disables clocksource verification at runtime, as well
6119 as the stability checks done at bootup. Used to enable
6120 high-resolution timer mode on older hardware, and in
6121 virtualized environment.
6122 [x86] noirqtime: Do not use TSC to do irq accounting.
6123 Used to run time disable IRQ_TIME_ACCOUNTING on any
6124 platforms where RDTSC is slow and this accounting
6126 [x86] unstable: mark the TSC clocksource as unstable, this
6127 marks the TSC unconditionally unstable at bootup and
6128 avoids any further wobbles once the TSC watchdog notices.
6129 [x86] nowatchdog: disable clocksource watchdog. Used
6130 in situations with strict latency requirements (where
6131 interruptions from clocksource watchdog are not
6134 tsc_early_khz= [X86] Skip early TSC calibration and use the given
6135 value instead. Useful when the early TSC frequency discovery
6136 procedure is not reliable, such as on overclocked systems
6137 with CPUID.16h support and partial CPUID.15h support.
6138 Format: <unsigned int>
6140 tsx= [X86] Control Transactional Synchronization
6141 Extensions (TSX) feature in Intel processors that
6142 support TSX control.
6144 This parameter controls the TSX feature. The options are:
6146 on - Enable TSX on the system. Although there are
6147 mitigations for all known security vulnerabilities,
6148 TSX has been known to be an accelerator for
6149 several previous speculation-related CVEs, and
6150 so there may be unknown security risks associated
6151 with leaving it enabled.
6153 off - Disable TSX on the system. (Note that this
6154 option takes effect only on newer CPUs which are
6155 not vulnerable to MDS, i.e., have
6156 MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get
6157 the new IA32_TSX_CTRL MSR through a microcode
6158 update. This new MSR allows for the reliable
6159 deactivation of the TSX functionality.)
6161 auto - Disable TSX if X86_BUG_TAA is present,
6162 otherwise enable TSX on the system.
6164 Not specifying this option is equivalent to tsx=off.
6166 See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6169 tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async
6170 Abort (TAA) vulnerability.
6172 Similar to Micro-architectural Data Sampling (MDS)
6173 certain CPUs that support Transactional
6174 Synchronization Extensions (TSX) are vulnerable to an
6175 exploit against CPU internal buffers which can forward
6176 information to a disclosure gadget under certain
6179 In vulnerable processors, the speculatively forwarded
6180 data can be used in a cache side channel attack, to
6181 access data to which the attacker does not have direct
6184 This parameter controls the TAA mitigation. The
6187 full - Enable TAA mitigation on vulnerable CPUs
6190 full,nosmt - Enable TAA mitigation and disable SMT on
6191 vulnerable CPUs. If TSX is disabled, SMT
6192 is not disabled because CPU is not
6193 vulnerable to cross-thread TAA attacks.
6194 off - Unconditionally disable TAA mitigation
6196 On MDS-affected machines, tsx_async_abort=off can be
6197 prevented by an active MDS mitigation as both vulnerabilities
6198 are mitigated with the same mechanism so in order to disable
6199 this mitigation, you need to specify mds=off too.
6201 Not specifying this option is equivalent to
6202 tsx_async_abort=full. On CPUs which are MDS affected
6203 and deploy MDS mitigation, TAA mitigation is not
6204 required and doesn't provide any additional
6208 Documentation/admin-guide/hw-vuln/tsx_async_abort.rst
6210 turbografx.map[2|3]= [HW,JOY]
6211 TurboGraFX parallel port interface
6213 <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7>
6214 See also Documentation/input/devices/joystick-parport.rst
6216 udbg-immortal [PPC] When debugging early kernel crashes that
6217 happen after console_init() and before a proper
6218 console driver takes over, this boot options might
6219 help "seeing" what's going on.
6221 uhash_entries= [KNL,NET]
6222 Set number of hash buckets for UDP/UDP-Lite connections
6225 [USB] Ignore overcurrent events (default N).
6226 Some badly-designed motherboards generate lots of
6227 bogus events, for ports that aren't wired to
6228 anything. Set this parameter to avoid log spamming.
6229 Note that genuine overcurrent events won't be
6233 [X86] Cause panic on unknown NMI.
6235 usbcore.authorized_default=
6236 [USB] Default USB device authorization:
6237 (default -1 = authorized except for wireless USB,
6238 0 = not authorized, 1 = authorized, 2 = authorized
6239 if device connected to internal port)
6241 usbcore.autosuspend=
6242 [USB] The autosuspend time delay (in seconds) used
6243 for newly-detected USB devices (default 2). This
6244 is the time required before an idle device will be
6245 autosuspended. Devices for which the delay is set
6246 to a negative value won't be autosuspended at all.
6248 usbcore.usbfs_snoop=
6249 [USB] Set to log all usbfs traffic (default 0 = off).
6251 usbcore.usbfs_snoop_max=
6252 [USB] Maximum number of bytes to snoop in each URB
6255 usbcore.blinkenlights=
6256 [USB] Set to cycle leds on hubs (default 0 = off).
6258 usbcore.old_scheme_first=
6259 [USB] Start with the old device initialization
6260 scheme (default 0 = off).
6262 usbcore.usbfs_memory_mb=
6263 [USB] Memory limit (in MB) for buffers allocated by
6264 usbfs (default = 16, 0 = max = 2047).
6266 usbcore.use_both_schemes=
6267 [USB] Try the other device initialization scheme
6268 if the first one fails (default 1 = enabled).
6270 usbcore.initial_descriptor_timeout=
6271 [USB] Specifies timeout for the initial 64-byte
6272 USB_REQ_GET_DESCRIPTOR request in milliseconds
6273 (default 5000 = 5.0 seconds).
6275 usbcore.nousb [USB] Disable the USB subsystem
6278 [USB] A list of quirk entries to augment the built-in
6279 usb core quirk list. List entries are separated by
6280 commas. Each entry has the form
6281 VendorID:ProductID:Flags. The IDs are 4-digit hex
6282 numbers and Flags is a set of letters. Each letter
6283 will change the built-in quirk; setting it if it is
6284 clear and clearing it if it is set. The letters have
6285 the following meanings:
6286 a = USB_QUIRK_STRING_FETCH_255 (string
6287 descriptors must not be fetched using
6289 b = USB_QUIRK_RESET_RESUME (device can't resume
6290 correctly so reset it instead);
6291 c = USB_QUIRK_NO_SET_INTF (device can't handle
6292 Set-Interface requests);
6293 d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't
6294 handle its Configuration or Interface
6296 e = USB_QUIRK_RESET (device can't be reset
6297 (e.g morph devices), don't use reset);
6298 f = USB_QUIRK_HONOR_BNUMINTERFACES (device has
6299 more interface descriptions than the
6300 bNumInterfaces count, and can't handle
6301 talking to these interfaces);
6302 g = USB_QUIRK_DELAY_INIT (device needs a pause
6303 during initialization, after we read
6304 the device descriptor);
6305 h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For
6306 high speed and super speed interrupt
6307 endpoints, the USB 2.0 and USB 3.0 spec
6308 require the interval in microframes (1
6309 microframe = 125 microseconds) to be
6310 calculated as interval = 2 ^
6312 Devices with this quirk report their
6313 bInterval as the result of this
6314 calculation instead of the exponent
6315 variable used in the calculation);
6316 i = USB_QUIRK_DEVICE_QUALIFIER (device can't
6317 handle device_qualifier descriptor
6319 j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device
6320 generates spurious wakeup, ignore
6321 remote wakeup capability);
6322 k = USB_QUIRK_NO_LPM (device can't handle Link
6324 l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL
6325 (Device reports its bInterval as linear
6326 frames instead of the USB 2.0
6328 m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs
6329 to be disconnected before suspend to
6330 prevent spurious wakeup);
6331 n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a
6332 pause after every control message);
6333 o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra
6334 delay after resetting its port);
6335 Example: quirks=0781:5580:bk,0a5c:5834:gij
6338 [USBHID] The interval which mice are to be polled at.
6341 [USBHID] The interval which joysticks are to be polled at.
6344 [USBHID] The interval which keyboards are to be polled at.
6346 usb-storage.delay_use=
6347 [UMS] The delay in seconds before a new device is
6348 scanned for Logical Units (default 1).
6351 [UMS] A list of quirks entries to supplement or
6352 override the built-in unusual_devs list. List
6353 entries are separated by commas. Each entry has
6354 the form VID:PID:Flags where VID and PID are Vendor
6355 and Product ID values (4-digit hex numbers) and
6356 Flags is a set of characters, each corresponding
6357 to a common usb-storage quirk flag as follows:
6358 a = SANE_SENSE (collect more than 18 bytes
6359 of sense data, not on uas);
6360 b = BAD_SENSE (don't collect more than 18
6361 bytes of sense data, not on uas);
6362 c = FIX_CAPACITY (decrease the reported
6363 device capacity by one sector);
6364 d = NO_READ_DISC_INFO (don't use
6365 READ_DISC_INFO command, not on uas);
6366 e = NO_READ_CAPACITY_16 (don't use
6367 READ_CAPACITY_16 command);
6368 f = NO_REPORT_OPCODES (don't use report opcodes
6370 g = MAX_SECTORS_240 (don't transfer more than
6371 240 sectors at a time, uas only);
6372 h = CAPACITY_HEURISTICS (decrease the
6373 reported device capacity by one
6374 sector if the number is odd);
6375 i = IGNORE_DEVICE (don't bind to this
6377 j = NO_REPORT_LUNS (don't use report luns
6379 k = NO_SAME (do not use WRITE_SAME, uas only)
6380 l = NOT_LOCKABLE (don't try to lock and
6381 unlock ejectable media, not on uas);
6382 m = MAX_SECTORS_64 (don't transfer more
6383 than 64 sectors = 32 KB at a time,
6385 n = INITIAL_READ10 (force a retry of the
6386 initial READ(10) command, not on uas);
6387 o = CAPACITY_OK (accept the capacity
6388 reported by the device, not on uas);
6389 p = WRITE_CACHE (the device cache is ON
6390 by default, not on uas);
6391 r = IGNORE_RESIDUE (the device reports
6392 bogus residue values, not on uas);
6393 s = SINGLE_LUN (the device has only one
6395 t = NO_ATA_1X (don't allow ATA(12) and ATA(16)
6396 commands, uas only);
6397 u = IGNORE_UAS (don't bind to the uas driver);
6398 w = NO_WP_DETECT (don't test whether the
6399 medium is write-protected).
6400 y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE
6401 even if the device claims no cache,
6403 Example: quirks=0419:aaf5:rl,0421:0433:rc
6405 user_debug= [KNL,ARM]
6407 See arch/arm/Kconfig.debug help text.
6408 1 - undefined instruction events
6410 4 - invalid data aborts
6413 Example: user_debug=31
6416 [X86] Flags controlling user PTE allocations.
6418 nohigh = do not allocate PTE pages in
6419 HIGHMEM regardless of setting
6422 vdso= [X86,SH,SPARC]
6423 On X86_32, this is an alias for vdso32=. Otherwise:
6425 vdso=1: enable VDSO (the default)
6426 vdso=0: disable VDSO mapping
6428 vdso32= [X86] Control the 32-bit vDSO
6429 vdso32=1: enable 32-bit VDSO
6430 vdso32=0 or vdso32=2: disable 32-bit VDSO
6432 See the help text for CONFIG_COMPAT_VDSO for more
6433 details. If CONFIG_COMPAT_VDSO is set, the default is
6434 vdso32=0; otherwise, the default is vdso32=1.
6436 For compatibility with older kernels, vdso32=2 is an
6439 Try vdso32=0 if you encounter an error that says:
6440 dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
6443 vector=percpu: enable percpu vector domain
6445 video= [FB] Frame buffer configuration
6446 See Documentation/fb/modedb.rst.
6448 video.brightness_switch_enabled= [ACPI]
6450 If set to 1, on receiving an ACPI notify event
6451 generated by hotkey, video driver will adjust brightness
6452 level and then send out the event to user space through
6453 the allocated input device. If set to 0, video driver
6454 will only send out the event without touching backlight
6459 [VMMIO] Memory mapped virtio (platform) device.
6461 <size>@<baseaddr>:<irq>[:<id>]
6463 <size> := size (can use standard suffixes
6465 <baseaddr> := physical base address
6466 <irq> := interrupt number (as passed to
6468 <id> := (optional) platform device id
6470 virtio_mmio.device=1K@0x100b0000:48:7
6472 Can be used multiple times for multiple devices.
6474 vga= [BOOT,X86-32] Select a particular video mode
6475 See Documentation/x86/boot.rst and
6476 Documentation/admin-guide/svga.rst.
6477 Use vga=ask for menu.
6478 This is actually a boot loader parameter; the value is
6479 passed to the kernel using a special protocol.
6481 vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y.
6482 May slow down system boot speed, especially when
6483 enabled on systems with a large amount of memory.
6484 All options are enabled by default, and this
6485 interface is meant to allow for selectively
6486 enabling or disabling specific virtual memory
6489 Available options are:
6490 P Enable page structure init time poisoning
6491 - Disable all of the above options
6493 vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact
6494 size of <nn>. This can be used to increase the
6495 minimum size (128MB on x86). It can also be used to
6496 decrease the size and leave more room for directly
6499 vmcp_cma=nn[MG] [KNL,S390]
6500 Sets the memory size reserved for contiguous memory
6501 allocations for the vmcp device driver.
6503 vmhalt= [KNL,S390] Perform z/VM CP command after system halt.
6506 vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic.
6509 vmpoff= [KNL,S390] Perform z/VM CP command after power off.
6513 Controls the behavior of vsyscalls (i.e. calls to
6514 fixed addresses of 0xffffffffff600x00 from legacy
6515 code). Most statically-linked binaries and older
6516 versions of glibc use these calls. Because these
6517 functions are at fixed addresses, they make nice
6518 targets for exploits that can control RIP.
6520 emulate [default] Vsyscalls turn into traps and are
6521 emulated reasonably safely. The vsyscall
6524 xonly Vsyscalls turn into traps and are
6525 emulated reasonably safely. The vsyscall
6526 page is not readable.
6528 none Vsyscalls don't work at all. This makes
6529 them quite hard to use for exploits but
6530 might break your system.
6532 vt.color= [VT] Default text color.
6533 Format: 0xYX, X = foreground, Y = background.
6534 Default: 0x07 = light gray on black.
6536 vt.cur_default= [VT] Default cursor shape.
6537 Format: 0xCCBBAA, where AA, BB, and CC are the same as
6538 the parameters of the <Esc>[?A;B;Cc escape sequence;
6539 see VGA-softcursor.txt. Default: 2 = underline.
6541 vt.default_blu= [VT]
6542 Format: <blue0>,<blue1>,<blue2>,...,<blue15>
6543 Change the default blue palette of the console.
6544 This is a 16-member array composed of values
6547 vt.default_grn= [VT]
6548 Format: <green0>,<green1>,<green2>,...,<green15>
6549 Change the default green palette of the console.
6550 This is a 16-member array composed of values
6553 vt.default_red= [VT]
6554 Format: <red0>,<red1>,<red2>,...,<red15>
6555 Change the default red palette of the console.
6556 This is a 16-member array composed of values
6562 Set system-wide default UTF-8 mode for all tty's.
6563 Default is 1, i.e. UTF-8 mode is enabled for all
6564 newly opened terminals.
6566 vt.global_cursor_default=
6569 Set system-wide default for whether a cursor
6570 is shown on new VTs. Default is -1,
6571 i.e. cursors will be created by default unless
6572 overridden by individual drivers. 0 will hide
6573 cursors, 1 will display them.
6575 vt.italic= [VT] Default color for italic text; 0-15.
6578 vt.underline= [VT] Default color for underlined text; 0-15.
6581 watchdog timers [HW,WDT] For information on watchdog timers,
6582 see Documentation/watchdog/watchdog-parameters.rst
6583 or other driver-specific files in the
6584 Documentation/watchdog/ directory.
6588 Set the hard lockup detector stall duration
6589 threshold in seconds. The soft lockup detector
6590 threshold is set to twice the value. A value of 0
6591 disables both lockup detectors. Default is 10
6594 workqueue.watchdog_thresh=
6595 If CONFIG_WQ_WATCHDOG is configured, workqueue can
6596 warn stall conditions and dump internal state to
6597 help debugging. 0 disables workqueue stall
6598 detection; otherwise, it's the stall threshold
6599 duration in seconds. The default value is 30 and
6600 it can be updated at runtime by writing to the
6601 corresponding sysfs file.
6603 workqueue.disable_numa
6604 By default, all work items queued to unbound
6605 workqueues are affine to the NUMA nodes they're
6606 issued on, which results in better behavior in
6607 general. If NUMA affinity needs to be disabled for
6608 whatever reason, this option can be used. Note
6609 that this also can be controlled per-workqueue for
6610 workqueues visible under /sys/bus/workqueue/.
6612 workqueue.power_efficient
6613 Per-cpu workqueues are generally preferred because
6614 they show better performance thanks to cache
6615 locality; unfortunately, per-cpu workqueues tend to
6616 be more power hungry than unbound workqueues.
6618 Enabling this makes the per-cpu workqueues which
6619 were observed to contribute significantly to power
6620 consumption unbound, leading to measurably lower
6621 power usage at the cost of small performance
6624 The default value of this parameter is determined by
6625 the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
6627 workqueue.debug_force_rr_cpu
6628 Workqueue used to implicitly guarantee that work
6629 items queued without explicit CPU specified are put
6630 on the local CPU. This guarantee is no longer true
6631 and while local CPU is still preferred work items
6632 may be put on foreign CPUs. This debug option
6633 forces round-robin CPU selection to flush out
6634 usages which depend on the now broken guarantee.
6635 When enabled, memory and cache locality will be
6638 x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of
6639 default x2apic cluster mode on platforms
6642 xen_512gb_limit [KNL,X86-64,XEN]
6643 Restricts the kernel running paravirtualized under Xen
6644 to use only up to 512 GB of RAM. The reason to do so is
6645 crash analysis tools and Xen tools for doing domain
6646 save/restore/migration must be enabled to handle larger
6649 xen_emul_unplug= [HW,X86,XEN]
6650 Unplug Xen emulated devices
6651 Format: [unplug0,][unplug1]
6652 ide-disks -- unplug primary master IDE devices
6653 aux-ide-disks -- unplug non-primary-master IDE devices
6654 nics -- unplug network devices
6655 all -- unplug all emulated devices (NICs and IDE disks)
6656 unnecessary -- unplugging emulated devices is
6657 unnecessary even if the host did not respond to
6659 never -- do not unplug even if version check succeeds
6661 xen_legacy_crash [X86,XEN]
6662 Crash from Xen panic notifier, without executing late
6663 panic() code such as dumping handler.
6665 xen_nopvspin [X86,XEN]
6666 Disables the qspinlock slowpath using Xen PV optimizations.
6667 This parameter is obsoleted by "nopvspin" parameter, which
6668 has equivalent effect for XEN platform.
6671 Disables the PV optimizations forcing the HVM guest to
6672 run as generic HVM guest with no PV drivers.
6673 This option is obsoleted by the "nopv" option, which
6674 has equivalent effect for XEN platform.
6676 xen_no_vector_callback
6677 [KNL,X86,XEN] Disable the vector callback for Xen
6678 event channel interrupts.
6680 xen_scrub_pages= [XEN]
6681 Boolean option to control scrubbing pages before giving them back
6682 to Xen, for use by other domains. Can be also changed at runtime
6683 with /sys/devices/system/xen_memory/xen_memory0/scrub_pages.
6684 Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT.
6686 xen_timer_slop= [X86-64,XEN]
6687 Set the timer slop (in nanoseconds) for the virtual Xen
6688 timers (default is 100000). This adjusts the minimum
6689 delta of virtualized Xen timers, where lower values
6690 improve timer resolution at the expense of processing
6691 more timer interrupts.
6693 xen.balloon_boot_timeout= [XEN]
6694 The time (in seconds) to wait before giving up to boot
6695 in case initial ballooning fails to free enough memory.
6696 Applies only when running as HVM or PVH guest and
6697 started with less memory configured than allowed at
6698 max. Default is 180.
6700 xen.event_eoi_delay= [XEN]
6701 How long to delay EOI handling in case of event
6702 storms (jiffies). Default is 10.
6704 xen.event_loop_timeout= [XEN]
6705 After which time (jiffies) the event handling loop
6706 should start to delay EOI handling. Default is 2.
6708 xen.fifo_events= [XEN]
6709 Boolean parameter to disable using fifo event handling
6710 even if available. Normally fifo event handling is
6711 preferred over the 2-level event handling, as it is
6712 fairer and the number of possible event channels is
6713 much higher. Default is on (use fifo events).
6715 nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
6716 Disables the PV optimizations forcing the guest to run
6717 as generic guest with no PV drivers. Currently support
6718 XEN HVM, KVM, HYPER_V and VMWARE guest.
6720 nopvspin [X86,XEN,KVM]
6721 Disables the qspinlock slow path using PV optimizations
6722 which allow the hypervisor to 'idle' the guest on lock
6725 xirc2ps_cs= [NET,PCMCIA]
6727 <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
6730 By default on POWER9 and above, the kernel will
6731 natively use the XIVE interrupt controller. This option
6732 allows the fallback firmware mode to be used:
6734 off Fallback to firmware control of XIVE interrupt
6735 controller on both pseries and powernv
6736 platforms. Only useful on POWER9 and above.
6738 xive.store-eoi=off [PPC]
6739 By default on POWER10 and above, the kernel will use
6740 stores for EOI handling when the XIVE interrupt mode
6741 is active. This option allows the XIVE driver to use
6742 loads instead, as on POWER9.
6744 xhci-hcd.quirks [USB,KNL]
6745 A hex value specifying bitmask with supplemental xhci
6746 host controller quirks. Meaning of each bit can be
6747 consulted in header drivers/usb/host/xhci.h.
6750 Format: { early | on | rw | ro | off }
6751 Controls if xmon debugger is enabled. Default is off.
6752 Passing only "xmon" is equivalent to "xmon=early".
6753 early Call xmon as early as possible on boot; xmon
6754 debugger is called from setup_arch().
6755 on xmon debugger hooks will be installed so xmon
6756 is only called on a kernel crash. Default mode,
6757 i.e. either "ro" or "rw" mode, is controlled
6758 with CONFIG_XMON_DEFAULT_RO_MODE.
6759 rw xmon debugger hooks will be installed so xmon
6760 is called only on a kernel crash, mode is write,
6761 meaning SPR registers, memory and, other data
6762 can be written using xmon commands.
6763 ro same as "rw" option above but SPR registers,
6764 memory, and other data can't be written using
6766 off xmon is disabled.